{"license":"CC-BY-4.0","count":154,"events":[{"id":"rss:pytorch:7f13355308fb","kind":"announcement","title":"Extending Inductor for the IBM Spyre Accelerator","headline":"PyTorch extiende su backend Inductor para el acelerador IBM Spyre","date":"2026-10-13","observed_at":"2026-10-13T12:54:15Z","subject":{"type":"feed","ref":"pytorch","url":"https://pytorch.org/blog/extending-inductor-for-the-ibm-spyre-accelerator/","entity":"pytorch","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["open-weights"],"affects":["pytorch"],"summary":{"why":"La migración del backend de compilación hacia Inductor mejora la compatibilidad y el rendimiento del stack de software de PyTorch en el acelerador IBM Spyre, alineándose con el ecosistema estándar de PyTorch."},"highlights":[{"text":"Migración del backend de compilación de sendnn a Inductor","quote":"moved its torch.compile backend from sendnn, a separate compilation stack built entirely from scratch, to backend=inductor","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T12:54:15Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://pytorch.org/blog/extending-inductor-for-the-ibm-spyre-accelerator/","title":"Extending Inductor for the IBM Spyre Accelerator","retrieved_at":"2026-10-13T12:53:56Z","sha256":"af6b6cfc519356ed772f7c5453b3ff53ba341e886681257dd64cffb380f00679","license":null},{"url":"https://pytorch.org/blog/extending-inductor-for-the-ibm-spyre-accelerator/","title":"Extending Inductor for the IBM Spyre Accelerator - PyTorch","retrieved_at":"2026-10-13T12:53:57Z","sha256":"6cd6c5cc51845a1469c95873db28e9d39268c099db7b32dcd32895a955f545a7","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--pytorch--7f13355308fb","related":[],"org":{"slug":"pytorch","name":"PyTorch","kind":"proyecto","url":"https://pytorch.org"}},{"id":"rss:xai-news:101d8ce4b445","kind":"announcement","title":"News: Research, Product & Company Updates | SpaceXSI","date":"2026-10-12","observed_at":"2026-10-13T10:08:48Z","subject":{"type":"feed","ref":"xai-news","url":"https://x.ai/news","entity":"xai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["open-weights"],"affects":["cursor"],"summary":{"why":"Se anuncia el lanzamiento de Grok 4.7, un modelo de nueva generación optimizado para tareas de programación y conocimiento, con mejoras significativas en velocidad y coste."},"highlights":[{"text":"Lanzamiento de Grok 4.7 para coding y conocimiento.","quote":"Introducing Grok 4.7 SpaceXSI's most powerful model for coding and knowledge work.","source":0},{"text":"Mejora de rendimiento y reducción de costes.","quote":"Twice as fast, at half the price of comparable models.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T10:08:48Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 2 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://x.ai/news","title":"News: Research, Product & Company Updates | SpaceXSI","retrieved_at":"2026-10-13T09:59:40Z","sha256":"f451256231f404fcf142c782583ea86f263afed4e7f33181b853aeef46578232","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--xai-news--101d8ce4b445","related":[],"org":{"slug":"xai","name":"xAI","kind":"laboratorio","url":"https://x.ai"}},{"id":"rss:xai-news:ec0714e7c8ff","kind":"announcement","title":"Introducing The X Subscription | SpaceXSI","date":"2026-10-12","observed_at":"2026-10-13T10:08:47Z","subject":{"type":"feed","ref":"xai-news","url":"https://x.ai/news/x-subscription","entity":"xai","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":[],"affects":["cursor"],"summary":{"why":"xAI integra Cursor en una nueva suscripción unificada, lo que cambia la forma en que se consume el uso de modelos Grok y agentes de codificación en Cursor."},"highlights":[{"text":"Unificación de servicios en un solo plan","quote":"Grok, Grok Bot, Cursor, and X Premium, all on one plan.","source":0},{"text":"Uso compartido de modelos Grok entre Grok y Cursor","quote":"Your shared pool covers Grok models whether you use them in the Grok apps, the Grok tab on X, or Cursor.","source":0},{"text":"Disponibilidad de modelos de terceros en Cursor","quote":"With the new plans, third-party models are still available in Cursor on Premium and above. Usage is billed separately at API list prices, through top-ups.","source":0},{"text":"Nuevos planes para equipos y empresas","quote":"Team users get Premium, Super, and Ultra on The X Subscription at the same per-seat prices as individual plans.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T10:08:47Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://x.ai/news/x-subscription","title":"Introducing The X Subscription | SpaceXSI","retrieved_at":"2026-10-13T10:04:53Z","sha256":"adbd43699ab08088a7f89b0b5d8cfac37ff16513d38dfbb0ed2795af53fcafd3","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--xai-news--ec0714e7c8ff","related":[],"org":{"slug":"xai","name":"xAI","kind":"laboratorio","url":"https://x.ai"}},{"id":"rss:google-gemini:a24fd97eb3d0","kind":"announcement","title":"Release notes | Gemini API | Google AI for Developers","date":"2026-10-12","observed_at":"2026-10-13T10:08:41Z","subject":{"type":"feed","ref":"google-gemini","url":"https://ai.google.dev/gemini-api/docs/changelog","entity":"google","channel":"n/a"},"status":"published","severity":"notable","action":"update","tags":[],"affects":[],"summary":{"why":"Se han lanzado múltiples modelos de nueva generación (Gemini 3.8 Flash, 3.7 Flash, 3.6 Flash) y se han anunciado deprecaciones críticas de versiones anteriores que requieren actualizar los model strings en las aplicaciones."},"highlights":[{"text":"Lanzamiento de Gemini 3.8 Flash para ingeniería de software y agentes.","quote":"Released gemini-3.8-flash , our most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows.","source":0},{"text":"Deprecación de Gemini 3.7 Flash y redirección automática a 3.8.","quote":"gemini-3.7-flash is deprecated and has been replaced by gemini-3.8-flash . It is no longer available to select in Google AI Studio, and API requests to gemini-3.7-flash are automatically routed to gemini-3.8-flash .","source":0},{"text":"Nuevos modelos de audio-a-audio (Live API) para aplicaciones de voz en tiempo real.","quote":"Released two new audio-to-audio models for real-time voice applications using the Live API: Gemini 3.8 Live ( gemini-3.8-live )","source":0},{"text":"Lanzamiento de Gemini Nano Banana 2.1 para generación de imágenes de alta eficiencia.","quote":"Released Gemini Nano Banana 2.1 ( gemini-nano-banana-2.1 ), the latest high-efficiency image generation and conversational editing model.","source":0},{"text":"Deprecación de parámetros de muestreo temperature, top_p y top_k.","quote":"The sampling parameters temperature , top_p and top_k are now deprecated.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T10:08:41Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://ai.google.dev/gemini-api/docs/changelog","title":"Release notes | Gemini API | Google AI for Developers","retrieved_at":"2026-10-13T09:59:40Z","sha256":"16c470f523fdf3599341cc7c8644ddd4cf42f3fe5d83359590ac66e977debd01","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--google-gemini--a24fd97eb3d0","related":[],"org":{"slug":"google","name":"Google","kind":"empresa","url":"https://deepmind.google"}},{"id":"github:ggml-org/llama.cpp:b11571","kind":"software_release","title":"llama.cpp b11571","date":"2026-10-12","observed_at":"2026-10-13T05:52:47Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11571","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"try","tags":["new-architecture"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":null},"highlights":[{"text":"Soporte para una nueva arquitectura de modelo","quote":"model: support vllm decision 2 model (#30378)","source":0},{"text":"Soporte para una nueva arquitectura de modelo","quote":"models: support pplx-decider","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:52:47Z","rejected":[]},"triage":{"method":"rules","model":"gemma4","reasons":["reglas: publish (new-architecture)","llm lo descartó pero hay señal fuerte; se aplica la decisión de reglas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11571","title":"ggml-org/llama.cpp b11571","retrieved_at":"2026-10-13T05:51:09Z","sha256":"e3fa221fa38b454624e9770f41733c80f36932fe401dd75c1bc6336cb178df61","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11571","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11571"},{"id":"github:ggml-org/llama.cpp:b11572","kind":"software_release","title":"llama.cpp b11572","date":"2026-10-12","observed_at":"2026-10-13T05:52:30Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11572","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"try","tags":["new-architecture"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":null},"highlights":[{"text":"Soporte para una nueva arquitectura de modelo","quote":"model: support vllm d3 models (#30383)","source":0},{"text":"Soporte para una nueva arquitectura de modelo","quote":"models: support pplx-decider","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:52:30Z","rejected":[]},"triage":{"method":"rules","model":"gemma4","reasons":["reglas: publish (new-architecture)","llm lo descartó pero hay señal fuerte; se aplica la decisión de reglas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11572","title":"ggml-org/llama.cpp b11572","retrieved_at":"2026-10-13T05:51:09Z","sha256":"066afdd3745aaeb0c6c06b209e76d9f06d574b76a5b0274da26ab5bf206865fc","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11572","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11572"},{"id":"github:ggml-org/llama.cpp:b11556","kind":"software_release","title":"llama.cpp b11556","date":"2026-10-11","observed_at":"2026-10-13T05:52:49Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11556","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"try","tags":["new-architecture"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":null},"highlights":[{"text":"Soporte para una nueva arquitectura de modelo","quote":"model :  support for Prism Bonsai 2 27B (#29600)","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:52:49Z","rejected":[]},"triage":{"method":"rules","model":"gemma4","reasons":["reglas: publish (new-architecture)","llm lo descartó pero hay señal fuerte; se aplica la decisión de reglas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11556","title":"ggml-org/llama.cpp b11556","retrieved_at":"2026-10-13T05:51:09Z","sha256":"579a5d9d465371744828e5cd195a8b959b60a078c5ffb2674ffffd66a51aba47","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11556","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11556"},{"id":"rss:zai-api:9156c37fb1c2","kind":"announcement","title":"New Released - Overview - Z.AI DEVELOPER DOCUMENT","date":"2026-10-10","observed_at":"2026-10-13T10:09:00Z","subject":{"type":"feed","ref":"zai-api","url":"https://docs.z.ai/release-notes/new-released","entity":"zai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["open-weights"],"affects":[],"summary":{"why":"El lanzamiento de la serie GLM-5 (incluyendo la versión Flash con capacidades visuales nativas y la versión estándar con mejoras en codificación) representa una nueva generación de modelos frontera con capacidades de agente y razonamiento avanzado para ingeniería de software."},"highlights":[{"text":"GLM-5.3-Flash introduce capacidades visuales nativas para observar interfaces y feedback de interacción.","quote":"Native visual capabilities enable the model to observe interfaces, rendering results, and interaction feedback—creating a closed loop across code, browsers, and GUIs.","source":0},{"text":"GLM-5.3 ofrece un aumento del 50% en capacidades de codificación respecto a la versión anterior.","quote":"GLM-5.3 delivers a significant improvement in coding capabilities, achieving a 50% gain over GLM-5.2 on Z.ai Code Bench","source":0},{"text":"GLM-5.3 incluye capacidades emergentes para revisión de código y descubrimiento de vulnerabilidades.","quote":"Emergent Cybersecurity Capabilities: GLM-5.3 matches Mythos 5 in white-box code review and vulnerability discovery.","source":0},{"text":"GLM-5.2 soporta un contexto de 1M de tokens sin pérdida.","quote":"Supports 1M lossless context, significantly improving long-horizon task capabilities and reducing context drift and goal forgetting in complex tasks","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T10:09:00Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://docs.z.ai/release-notes/new-released","title":"New Released - Overview - Z.AI DEVELOPER DOCUMENT","retrieved_at":"2026-10-13T09:59:42Z","sha256":"5df86ed3419d305fa6aacf38466cbfa484dce0ee5d452b91f6e09afd7442592b","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--zai-api--9156c37fb1c2","related":["hf:zai-org/GLM-5.3"],"org":{"slug":"zai","name":"Z.ai (Zhipu)","kind":"laboratorio","url":"https://z.ai"}},{"id":"rss:xai-api:2b950d5a745e","kind":"announcement","title":"Release Notes | SpaceXSI Docs","date":"2026-10-10","observed_at":"2026-10-13T10:08:52Z","subject":{"type":"feed","ref":"xai-api","url":"https://docs.x.ai/developers/release-notes","entity":"xai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":[],"affects":["cursor"],"summary":{"why":"El lanzamiento de Grok 4.7 introduce un nuevo modelo frontera con ventana de contexto de 500k y capacidades de razonamiento configurable, además de cambios en la estructura de precios y soporte para visión."},"highlights":[{"text":"Lanzamiento de Grok 4.7","quote":"Grok 4.7, SpaceXSI's frontier model for coding, agentic tasks, and knowledge work, is now available on the xAI API as grok-4.7 .","source":0},{"text":"Ventana de contexto de 500k y capacidades multimodales","quote":"It has a 500k context window, text and image inputs with text-only output, and no text output limit.","source":0},{"text":"Esfuerzo de razonamiento configurable","quote":"Reasoning effort supports low, medium, high (default), and xhigh.","source":0},{"text":"Nueva estructura de precios por volumen de tokens","quote":"Pricing is $2 / $0.50 / $6 per 1M tokens (input / cached input / output) below 200k prompt tokens, and $4 / $1 / $12 above.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T10:08:52Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://docs.x.ai/developers/release-notes","title":"Release Notes | SpaceXSI Docs","retrieved_at":"2026-10-13T09:59:41Z","sha256":"906fc3bc7f380fa611dbfc2aed77697694664e606c2c3934436ee6e54818c0a1","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--xai-api--2b950d5a745e","related":[],"org":{"slug":"xai","name":"xAI","kind":"laboratorio","url":"https://x.ai"}},{"id":"hf:nvidia/agile_one_s_pick_ssd_n2_1120","kind":"model_release","title":"nvidia publica agile_one_s_pick_ssd_n2_1120","date":"2026-10-10","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"nvidia/agile_one_s_pick_ssd_n2_1120","url":"https://huggingface.co/nvidia/agile_one_s_pick_ssd_n2_1120","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"created_at","value":"2026-10-10T00:20:17.000Z","evidence":{"source":0,"locator":"createdAt"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/nvidia/agile_one_s_pick_ssd_n2_1120","title":"Hugging Face · nvidia/agile_one_s_pick_ssd_n2_1120","retrieved_at":"2026-10-13T08:55:00Z","sha256":"e3bcc7f1feaa9a01a5504cb73ed28ccd000a385c6c2df6832f37101c2813c8a8","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"nvidia/agile_one_s_pick_ssd_n2_1120","params_total":null,"architecture":null,"context_length":null,"license":null,"gated":false,"pipeline":null,"variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--nvidia--agile_one_s_pick_ssd_n2_1120","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}},{"id":"hf:nvidia/agile_one_s_place_ssd_n2_1260","kind":"model_release","title":"nvidia publica agile_one_s_place_ssd_n2_1260","date":"2026-10-10","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"nvidia/agile_one_s_place_ssd_n2_1260","url":"https://huggingface.co/nvidia/agile_one_s_place_ssd_n2_1260","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"created_at","value":"2026-10-10T00:18:56.000Z","evidence":{"source":0,"locator":"createdAt"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/nvidia/agile_one_s_place_ssd_n2_1260","title":"Hugging Face · nvidia/agile_one_s_place_ssd_n2_1260","retrieved_at":"2026-10-13T08:55:00Z","sha256":"c92c4719f08324d3e434080a0a839aa7d3bb4a8f19aee3eb88b7b24bc8b2a64c","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"nvidia/agile_one_s_place_ssd_n2_1260","params_total":null,"architecture":null,"context_length":null,"license":null,"gated":false,"pipeline":null,"variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--nvidia--agile_one_s_place_ssd_n2_1260","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}},{"id":"github:ggml-org/llama.cpp:b11551","kind":"software_release","title":"llama.cpp b11551","date":"2026-10-10","observed_at":"2026-10-13T05:52:54Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11551","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"try","tags":["new-architecture"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":null},"highlights":[{"text":"Soporte para una nueva arquitectura de modelo","quote":"model : support MiniCPM-V 4.7 (#29416)","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:52:54Z","rejected":[]},"triage":{"method":"rules","model":"gemma4","reasons":["reglas: publish (new-architecture)","llm lo descartó pero hay señal fuerte; se aplica la decisión de reglas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11551","title":"ggml-org/llama.cpp b11551","retrieved_at":"2026-10-13T05:51:09Z","sha256":"f47c1ffc941e91345672ca8a86e53d4e65dce0a86099db872826fbe4a3b11639","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11551","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11551"},{"id":"github:open-webui/open-webui:v0.12.0","kind":"software_release","title":"open-webui v0.12.0","date":"2026-10-10","observed_at":"2026-10-12T21:07:42Z","subject":{"type":"repo","ref":"open-webui/open-webui","url":"https://github.com/open-webui/open-webui/releases/tag/v0.12.0","entity":"open-webui","channel":"auto-build"},"status":"published","severity":"notable","action":"try","tags":["feature","new-model","security"],"affects":["open-webui"],"summary":{"why":"Esta actualización introduce capacidades críticas de seguridad (MFA), mejoras significativas en la gestión de contexto para herramientas (Tool Search) y soporte para llamadas de voz en tiempo real."},"highlights":[{"text":"Implementación de autenticación de dos factores (MFA).","quote":"Administrators can require an authenticator app for every user under Admin Settings > Authentication, or with \"ENABLE_MFA\"","source":0},{"text":"Búsqueda de herramientas para optimizar el uso de tokens y caché.","quote":"Administrators can now turn on Tool Search, marked experimental, in Admin Settings > Interface, so tools with long definitions are left out of the request and only listed by name and a short description, and the model looks up the full definition with a `search_tools` tool when it needs one","source":0},{"text":"Soporte para llamadas de voz en tiempo real con modelos OpenAI.","quote":"Administrators can now switch Call mode in the Voice calls section of Audio settings from Standard to Realtime, so Voice mode talks through an OpenAI Realtime voice model","source":0},{"text":"Compacción de contexto durante ejecuciones largas de herramientas.","quote":"With context compaction on and native function calling, a reply that runs many tool calls in a row is now compacted between tool rounds once its context passes the token threshold","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-12T21:07:42Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (security)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/open-webui/open-webui/releases/tag/v0.12.0","title":"open-webui/open-webui v0.12.0","retrieved_at":"2026-10-12T21:02:29Z","sha256":"e5263d37e115f5719e76c06dc7d8a672e5354848ef07ceb0b27f1796eef70d10","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--open-webui--open-webui--v0.12.0","related":[],"org":{"slug":"open-webui","name":"Open WebUI","kind":"proyecto","url":"https://openwebui.com"},"tool":{"slug":"open-webui","name":"Open WebUI"},"version":"v0.12.0"},{"id":"rss:cohere-changelog:bd8460f367dc","kind":"announcement","title":"Release Notes | Cohere","date":"2026-10-09","observed_at":"2026-10-13T10:09:07Z","subject":{"type":"feed","ref":"cohere-changelog","url":"https://docs.cohere.com/changelog","entity":"cohere","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":[],"affects":[],"summary":{"why":"Cohere lanza su nueva familia de modelos de embeddings con capacidades multimodales (texto e imagen) y una ventana de contexto de 128k, permitiendo mejorar la calidad de recuperación en RAG y búsqueda semántica."},"highlights":[{"text":"Lanzamiento de la familia de modelos Embed 5 con dos variantes: Pro y Fast.","quote":"Announcing Cohere's Embed 5 Models We’re pleased to announce the release of Embed 5 , Cohere’s most powerful embeddings family yet.","source":0},{"text":"Capacidad de procesar entradas multimodales incluyendo imágenes y texto.","quote":"Multimodal inputs : Embed text, images, and mixed text-and-image inputs (e.g. PDF pages) in a single vector","source":0},{"text":"Soporte para una ventana de contexto extendida de 128k tokens.","quote":"Extended context length : 128k token context window","source":0},{"text":"Implementación de Matryoshka embeddings para dimensiones flexibles.","quote":"Flexible storage : Matryoshka embeddings in the following dimensions: [256, 512, 768, 1024, 1536, 2048] , with float , int8 , and binary output types","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T10:09:07Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://docs.cohere.com/changelog","title":"Release Notes | Cohere","retrieved_at":"2026-10-13T09:59:43Z","sha256":"78b1109dcc36358815476fbb2ddc88bfa7e4c0b24de437c3629e97ba14345f1d","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--cohere-changelog--bd8460f367dc","related":[],"org":{"slug":"cohere","name":"Cohere","kind":"laboratorio","url":"https://cohere.com"}},{"id":"rss:anthropic-news:c4a1f46e5a98","kind":"announcement","title":"Introducing the Anthropic Cyber Mission \\ Anthropic","date":"2026-10-08","observed_at":"2026-10-13T10:08:19Z","subject":{"type":"feed","ref":"anthropic-news","url":"https://www.anthropic.com/news/anthropic-cyber-mission","entity":"anthropic","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["open-weights"],"affects":["transformers"],"summary":{"why":"Anthropic lanza una herramienta de escaneo de vulnerabilidades para proyectos open-source (OSS Scanner) que utiliza sus modelos más capaces para generar reportes, pruebas de concepto y sugerencias de parches de forma automatizada."},"highlights":[{"text":"Lanzamiento de OSS Scanner para proyectos de código abierto.","quote":"Today, we’re launching OSS Scanner , which offers open-source projects regular security scans from our strongest models, for free.","source":0},{"text":"Los reportes incluyen pruebas de concepto y sugerencias de corrección.","quote":"Each report includes a proof of concept of how the bug could be exploited, an explanation, and a suggested fix where one is available.","source":0},{"text":"El proceso es totalmente automatizado y sin revisión humana previa.","quote":"The reports are model-generated and sent without human review.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T10:08:19Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 3 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://www.anthropic.com/news/anthropic-cyber-mission","title":"Introducing the Anthropic Cyber Mission \\ Anthropic","retrieved_at":"2026-10-13T10:04:34Z","sha256":"e909b979318a67b677a2150f526a70aa9c207cbca63c539084024490d280011d","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--anthropic-news--c4a1f46e5a98","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"}},{"id":"hf:mistralai/Voxtral-Mini-4B-Realtime-Arabic","kind":"model_release","title":"mistralai publica Voxtral-Mini-4B-Realtime-Arabic","date":"2026-10-08","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"mistralai/Voxtral-Mini-4B-Realtime-Arabic","url":"https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-Arabic","entity":"mistralai","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 4.4 B parámetros, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":4429679360,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"automatic-speech-recognition","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-10-08T11:36:06.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"VoxtralRealtimeForConditionalGeneration","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":131072,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-Arabic","title":"Hugging Face · mistralai/Voxtral-Mini-4B-Realtime-Arabic","retrieved_at":"2026-10-13T08:54:53Z","sha256":"d27efcabaabe89888e33211cf0f1d05a11989f1fe1d75ffee19e2b7c4063fae4","license":null},{"url":"https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-Arabic/blob/main/config.json","title":"mistralai/Voxtral-Mini-4B-Realtime-Arabic · config.json","retrieved_at":"2026-10-13T08:54:53Z","sha256":"a4a09e9d258cf9847ee7a6a31cabe7fb2d4b0cf660f15d26108d85c30be77098","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"mistralai/Voxtral-Mini-4B-Realtime-Arabic","params_total":4429679360,"architecture":"VoxtralRealtimeForConditionalGeneration","context_length":131072,"license":null,"gated":false,"pipeline":"automatic-speech-recognition","variants":[],"kv_bytes_per_token":106496,"quants":[]},"slug":"hf--mistralai--voxtral-mini-4b-realtime-arabic","related":[],"org":{"slug":"mistral","name":"Mistral AI","kind":"laboratorio","url":"https://mistral.ai"}},{"id":"github:ggml-org/llama.cpp:b11483","kind":"software_release","title":"llama.cpp b11483","date":"2026-10-08","observed_at":"2026-10-13T05:53:07Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11483","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"try","tags":["new-architecture"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":null},"highlights":[{"text":"Soporte para una nueva arquitectura de modelo","quote":"model : add LiquidAI/d1-3B decision model (#30110)","source":0},{"text":"Soporte para una nueva arquitectura de modelo","quote":"model : add LiquidAI/d1-3b decision model","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:53:07Z","rejected":[]},"triage":{"method":"rules","model":"gemma4","reasons":["reglas: publish (new-architecture)","llm lo descartó pero hay señal fuerte; se aplica la decisión de reglas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11483","title":"ggml-org/llama.cpp b11483","retrieved_at":"2026-10-13T05:51:10Z","sha256":"40e09a05834083eb83cfac82c036916e64bd254feae5c9d2465442019c8056d5","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11483","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11483"},{"id":"github:ggml-org/llama.cpp:b11488","kind":"software_release","title":"llama.cpp b11488","date":"2026-10-08","observed_at":"2026-10-13T05:53:04Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11488","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"try","tags":["new-architecture"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":null},"highlights":[{"text":"Soporte para una nueva arquitectura de modelo","quote":"model : add LiquidAI/d1-omni-600M decision model (#30114)","source":0},{"text":"Soporte para una nueva arquitectura de modelo","quote":"model : add LiquidAI/d1-omni-600M decision model","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:53:04Z","rejected":[]},"triage":{"method":"rules","model":"gemma4","reasons":["reglas: publish (new-architecture)","llm lo descartó pero hay señal fuerte; se aplica la decisión de reglas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11488","title":"ggml-org/llama.cpp b11488","retrieved_at":"2026-10-13T05:51:10Z","sha256":"3a0fabb05b305f399c45f89768707f151c387edfcac8ca54cae698f3fb6a5530","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11488","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11488"},{"id":"github:ggml-org/llama.cpp:b11495","kind":"software_release","title":"llama.cpp b11495","date":"2026-10-08","observed_at":"2026-10-13T05:52:57Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11495","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"try","tags":["new-architecture"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":null},"highlights":[{"text":"Soporte para una nueva arquitectura de modelo","quote":"model : support classifier_activation for rerankers (#29692)","source":0},{"text":"Soporte para una nueva arquitectura de modelo","quote":"model : support classifier_activation for rerankers","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:52:57Z","rejected":[]},"triage":{"method":"rules","model":"gemma4","reasons":["reglas: publish (new-architecture)","llm lo descartó pero hay señal fuerte; se aplica la decisión de reglas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11495","title":"ggml-org/llama.cpp b11495","retrieved_at":"2026-10-13T05:51:10Z","sha256":"41a69f8cb64b49348c1814a10574972b53fad58a97e760ff503a463b318bc05e","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11495","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11495"},{"id":"rss:pytorch:b1835280789a","kind":"announcement","title":"Building Spyre as a Native PyTorch Device","date":"2026-10-08","observed_at":"2026-10-12T21:09:30Z","subject":{"type":"feed","ref":"pytorch","url":"https://pytorch.org/blog/building-spyre-as-a-native-pytorch-device/","entity":"pytorch","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-architecture"],"affects":["pytorch"],"summary":{"why":"Permite que Spyre se integre directamente en el ecosistema de PyTorch mediante abstracciones nativas, facilitando su uso como un dispositivo de cómputo estándar."},"highlights":[{"text":"Spyre se convierte en un dispositivo nativo de PyTorch conectando sus abstracciones existentes al runtime y firmware de Spyre.","quote":"Spyre becomes a native PyTorch device by connecting PyTorch’s existing device, allocator, stream, and compiler abstractions through torch-spyre to the Spyre runtime and firmware.","source":0},{"text":"Uso de PrivateUse1 para dar capacidades reales de dispositivo.","quote":"PrivateUse1 gives Spyre a real...","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-12T21:09:30Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 2 citas verificadas, 0 rechazadas","publish con severidad minor: queda en observación","corrección editorial: lanzamiento real nunca es minor (la severidad minor era un fallo de calibración del prompt)"]},"sources":[{"url":"https://pytorch.org/blog/building-spyre-as-a-native-pytorch-device/","title":"Building Spyre as a Native PyTorch Device","retrieved_at":"2026-10-12T21:02:37Z","sha256":"74c4e3136ec669e6f66cc1eff38df7010152857daa60fca7e03df583b29e126e","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--pytorch--b1835280789a","related":[],"org":{"slug":"pytorch","name":"PyTorch","kind":"proyecto","url":"https://pytorch.org"}},{"id":"github:anthropics/claude-code:v2.1.295","kind":"software_release","title":"claude-code v2.1.295","date":"2026-10-08","observed_at":"2026-10-12T21:06:03Z","subject":{"type":"repo","ref":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.295","entity":"anthropic","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["fix","performance","security"],"affects":["claude-code"],"summary":{"why":"Esta actualización incluye mejoras críticas en la gestión de errores de red/timeouts para gateways (Bedrock, Vertex), correcciones de estabilidad en conexiones MCP y parches de seguridad/comportamiento en el manejo de herramientas y plugins."},"highlights":[{"text":"Soporte para timeouts en gateways de nube (Bedrock, Vertex, etc.)","quote":"Added support for `timeouts.upstream_ttfb_ms` on the Claude apps gateway's Bedrock, Vertex, Foundry and other cloud upstreams","source":0},{"text":"Corrección de errores de contexto en modelos de 1M de tokens","quote":"Fixed every request failing on a `[1m]` model when a gateway, Bedrock, Vertex or Foundry refuses the context-1m beta; Claude Code now resends without it","source":0},{"text":"Mejoras en la estabilidad de reconexión de servidores MCP remotos","quote":"Fixed remote MCP servers in headless and SDK sessions staying disconnected after an outage longer than 15 seconds, or reconnecting in a tight loop","source":0},{"text":"Corrección de errores de guardado de archivos por herramientas MCP","quote":"Fixed CSS, JavaScript and XML files returned by MCP tools being saved as .bin, which the Read tool refuses","source":0},{"text":"Control de reintentos para errores 429 y 529","quote":"Added `CLAUDE_CODE_RETRY_WATCHDOG_MAX_WAIT_MS` to limit how long unattended retry mode (`CLAUDE_CODE_RETRY_WATCHDOG`) waits out 429 and 529 errors","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-12T21:06:03Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (deprecation, feature)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.295","title":"anthropics/claude-code v2.1.295","retrieved_at":"2026-10-12T21:02:13Z","sha256":"287c6196d33a464561d0d47da9a708b61600d56cf543207ba2b130b99cb9adf2","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--anthropics--claude-code--v2.1.295","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"},"tool":{"slug":"claude-code","name":"Claude Code"},"version":"v2.1.295"},{"id":"github:ollama/ollama:v0.40.2","kind":"software_release","title":"ollama v0.40.2","date":"2026-10-08","observed_at":"2026-10-12T21:05:14Z","subject":{"type":"repo","ref":"ollama/ollama","url":"https://github.com/ollama/ollama/releases/tag/v0.40.2","entity":"ollama","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["breaking","performance"],"affects":["ollama"],"summary":{"why":"Introduce una actualización automática de modelos para mejorar el rendimiento y la compatibilidad con llama.cpp, lo que consume espacio adicional en disco debido a la creación de backups."},"highlights":[{"text":"Mejora de rendimiento y compatibilidad mediante la actualización de modelos en segundo plano para llama.cpp.","quote":"Models downloaded with earlier versions of Ollama are upgraded in the background the first time you run them, for better performance and compatibility when running on llama.cpp.","source":0},{"text":"Los modelos actualizados mantienen una copia de seguridad en disco para permitir downgrades seguros.","quote":"To make downgrading safe, Ollama keeps the original copy as a backup, so upgraded models are temporarily kept on disk.","source":0},{"text":"Corrección en la visualización de modelos duplicados al listar.","quote":"ollama list no longer shows duplicate entries for upgraded models","source":0},{"text":"Uso del contexto completo al lanzar modelos específicos.","quote":"ollama launch claude uses the model's full context length","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-12T21:05:14Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: drop (sin señales)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/ollama/ollama/releases/tag/v0.40.2","title":"ollama/ollama v0.40.2","retrieved_at":"2026-10-12T21:02:10Z","sha256":"bc477da4acc2913c04e18cf3bbdf936f40aa7f1604a77240763cdd660d614859","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ollama--ollama--v0.40.2","related":[],"org":{"slug":"ollama","name":"Ollama","kind":"proyecto","url":"https://ollama.com"},"tool":{"slug":"ollama","name":"Ollama"},"version":"v0.40.2"},{"id":"github:ggml-org/llama.cpp:b11509","kind":"software_release","title":"llama.cpp b11509","date":"2026-10-08","observed_at":"2026-10-12T21:02:57Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11509","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"update","tags":[],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Corrige un error de comparación de versiones en CUDA que causaba una pérdida de rendimiento silenciosa (fallback a un camino más lento) al actualizar a versiones mayores de CCCL (ej. 4.x o 5.x). Los usuarios con CUDA reciente pueden notar una regresión de rendimiento si no actualizan."},"highlights":[{"text":"Se corrige la lógica de guardia de versión de CCCL que fallaba al cambiar de mayor versión (ej. de 3.x a 4.x), evitando un fallback silencioso a un camino de ejecución menos eficiente.","quote":"Minor resets to 0 whenever a new major series is cut, so on CCCL 4.x this evaluates as 4 >= 3 && 0 >= 1, i.e. false. STRIDED_ITERATOR_AVAILABLE stops being defined and argsort silently falls back to the init_offsets path.","source":0}],"facts":[],"verification":{"status":"partial","checked_at":"2026-10-12T21:02:57Z","rejected":["CCCL already exposes the version as a single packed integer in MMMmmmpp form... so 3.4.3 is 3004003 and \">= 3.1\" is a plain \">= 3001000\".","Ubuntu x64 (CUDA 13)... Ubuntu x64 (ROCm 10.0)"]},"triage":{"method":"llm","model":"qwen3.6","reasons":["reglas: publish (breaking)","llm: publish; 1 citas verificadas, 2 rechazadas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11509","title":"ggml-org/llama.cpp b11509","retrieved_at":"2026-10-12T21:02:06Z","sha256":"0f5cbcb8d6332f326547adf9f9d7f09a9c4ca507bdcd381644485022cb0f586d","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11509","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11509"},{"id":"rss:openai:a097ef0d441a","kind":"announcement","title":"GPT-6 and Intelligent UI for everyone","date":"2026-10-07","observed_at":"2026-10-13T09:05:47Z","subject":{"type":"feed","ref":"openai","url":"https://openai.com/index/gpt-6-for-everyone","entity":"openai","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":[],"affects":[],"summary":{"why":"El lanzamiento de una nueva generación de modelo frontera (GPT-6) cambia el estado del arte y las capacidades que un desarrollador debe evaluar para sus aplicaciones."},"highlights":[{"text":"Lanzamiento global de GPT-6 con interfaz inteligente e interactiva.","quote":"GPT‑6 is rolling out globally in ChatGPT with Intelligent UI","source":0},{"text":"Respuestas más rápidas con capacidades visuales.","quote":"delivering faster responses with visuals and interactive experiences","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T09:05:47Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 2 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://openai.com/index/gpt-6-for-everyone","title":"GPT-6 and Intelligent UI for everyone","retrieved_at":"2026-10-13T09:05:01Z","sha256":"a02554a96319179dd58ff0353c63a4f79719fcf4e8bbe4938509d11f5ebd6ca9","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--openai--a097ef0d441a","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"}},{"id":"rss:hf-blog:16b7810afb00","kind":"announcement","title":"Introducing Falcon ASR","date":"2026-10-07","observed_at":"2026-10-13T06:17:57Z","subject":{"type":"feed","ref":"hf-blog","url":"https://huggingface.co/blog/tiiuae/falcon-asr","entity":"huggingface","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["feature","new-model"],"affects":["pytorch","transformers"],"summary":{"why":"Se introduce Falcon ASR, un nuevo modelo especializado en reconocimiento automático de voz (Automatic Speech Recognition)."},"highlights":[{"text":"Lanzamiento de Falcon ASR","quote":"Introducing Falcon ASR","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:17:57Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas","publish con severidad minor: queda en observación","corrección editorial: lanzamiento real nunca es minor (la severidad minor era un fallo de calibración del prompt)"]},"sources":[{"url":"https://huggingface.co/blog/tiiuae/falcon-asr","title":"Introducing Falcon ASR","retrieved_at":"2026-10-13T05:51:54Z","sha256":"38295e9cb6eb62bbb2fed493ff47f4686ec9b562238852f106beb8c01dd62db5","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--hf-blog--16b7810afb00","related":[],"org":{"slug":"huggingface","name":"Hugging Face","kind":"empresa","url":"https://huggingface.co"}},{"id":"github:turboderp-org/exllamav3:v1.6.0","kind":"software_release","title":"exllamav3 1.6.0","date":"2026-10-07","observed_at":"2026-10-12T21:07:38Z","subject":{"type":"repo","ref":"turboderp-org/exllamav3","url":"https://github.com/turboderp-org/exllamav3/releases/tag/v1.6.0","entity":"exllama","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["hardware","performance"],"affects":["exllamav3"],"summary":{"why":"Introduce soporte preliminar para ROCm y optimizaciones significativas en la gestión de CPU y memoria, mejorando el rendimiento en hardware específico."},"highlights":[{"text":"Soporte preliminar para ROCm (gfx1100+)","quote":"Preliminary ROCm support (ROCm 10 only, gfx1100+)","source":0},{"text":"Offloading más rápido para CPUs con AVX2","quote":"Faster offloading for AVX2 CPUs","source":0},{"text":"Reducción del uso de memoria del sistema","quote":"Reduced system memory overhead","source":0},{"text":"Corrección de vision towers para modelos GLM4 y Qwen3-VL","quote":"Fix vision towers for GLM4, Qwen3-VL","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-12T21:07:38Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: drop (sin señales)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/turboderp-org/exllamav3/releases/tag/v1.6.0","title":"turboderp-org/exllamav3 1.6.0","retrieved_at":"2026-10-12T21:02:28Z","sha256":"29f43754211be6b059c5b70bc89939adab8698d2c2818e38155c9b6b56b53f4d","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--turboderp-org--exllamav3--v1.6.0","related":[],"org":{"slug":"exllama","name":"ExLlama","kind":"proyecto","url":"https://github.com/turboderp-org/exllamav3"},"tool":{"slug":"exllamav3","name":"ExLlamaV3"},"version":"v1.6.0"},{"id":"github:openai/codex:rust-v0.161.0","kind":"software_release","title":"codex 0.161.0","date":"2026-10-07","observed_at":"2026-10-12T21:07:10Z","subject":{"type":"repo","ref":"openai/codex","url":"https://github.com/openai/codex/releases/tag/rust-v0.161.0","entity":"openai","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":[],"affects":["codex"],"summary":{"why":"Esta actualización introduce cambios significativos en el modelo por defecto (GPT-6.1 Sol), nuevas capacidades de razonamiento ultra en Amazon Bedrock y mejoras críticas en la gestión de permisos de archivos y sandboxing en Windows."},"highlights":[{"text":"Cambio del modelo por defecto a GPT-6.1 Sol.","quote":"GPT-6.1 Sol is now the default model in the bundled and Amazon Bedrock catalogs.","source":0},{"text":"Soporte para razonamiento Ultra y multi-agente en Amazon Bedrock.","quote":"Amazon Bedrock supports multi-agent V2 and Ultra reasoning on compatible models","source":0},{"text":"Mejoras en la escalada de permisos de sistema de archivos.","quote":"Approved filesystem escalation can now grant broader write access while preserving denied reads and network restrictions.","source":0},{"text":"Nuevas funciones de login para servidores MCP desde la terminal.","quote":"Sign in to MCP servers from an active terminal session with `/mcp login <name>`.","source":0},{"text":"Mejoras en el sandboxing de PowerShell para Windows.","quote":"sandboxed PowerShell preserves relative paths beneath protected user profiles.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-12T21:07:10Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: drop (sin señales)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/openai/codex/releases/tag/rust-v0.161.0","title":"openai/codex 0.161.0","retrieved_at":"2026-10-12T21:02:21Z","sha256":"1682643a811c3d1d8c6cd1dcae1108077777cc9bc9127960c8c54bed3b43ec14","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--openai--codex--rust-v0.161.0","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"},"tool":{"slug":"codex","name":"Codex CLI"},"version":"rust-v0.161.0"},{"id":"github:anthropics/claude-code:v2.1.293","kind":"software_release","title":"claude-code v2.1.293","date":"2026-10-07","observed_at":"2026-10-12T21:06:14Z","subject":{"type":"repo","ref":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.293","entity":"anthropic","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["fix","new-model","performance"],"affects":["claude-code"],"summary":{"why":"Introduce el nuevo modelo Claude Haiku 5.5 por defecto y corrige fugas de memoria en conexiones MCP y errores críticos de gestión de sesiones y herramientas."},"highlights":[{"text":"Soporte para Claude Haiku 5.5 como modelo predeterminado","quote":"Added Claude Haiku 5.5 (`claude-haiku-5-5`), now the default Haiku model on the Anthropic API","source":0},{"text":"Corrección de fuga de memoria en conexiones HTTP MCP","quote":"Fixed a memory leak where an HTTP MCP connection kept every request it had sent until it closed","source":0},{"text":"Mejora en la gestión de herramientas para mods","quote":"Added `isDeferred` to `$.tool.register` for mods: `false` lists the tool's schema in the prompt from the start instead of behind tool search","source":0},{"text":"Corrección de errores en la ejecución de comandos Bash y lectura de archivos","quote":"Fixed path-scoped rules and nested CLAUDE.md files not loading when Claude views a file with a single-file cat, head, tail, sed -n or grep command in the Bash tool instead of the Read tool","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-12T21:06:14Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (feature)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.293","title":"anthropics/claude-code v2.1.293","retrieved_at":"2026-10-12T21:02:13Z","sha256":"f152bef41559759b3cbdf8051b3c75b2f09fb54d483555eecda0d64aba63aa81","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--anthropics--claude-code--v2.1.293","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"},"tool":{"slug":"claude-code","name":"Claude Code"},"version":"v2.1.293"},{"id":"github:ggml-org/llama.cpp:b11474","kind":"software_release","title":"llama.cpp b11474","date":"2026-10-07","observed_at":"2026-10-12T21:04:41Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11474","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"update","tags":["new-model","performance"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Añade soporte para la arquitectura GLM5-Next con Multi-Token Prediction (MTP) y optimiza significativamente el rendimiento de los kernels para este tipo de modelos."},"highlights":[{"text":"Soporte para GLM5-Next MTP","quote":"add GLM5-Next NextN (MTP) graph","source":0},{"text":"Optimización de kernels para el catch-up de tokens","quote":"The 4-token catch-up falls from 6.9 ms to 0.33 ms of kernels","source":0},{"text":"Carga de archivos GGUF divididos","quote":"Make the trunk tensors optional when the file only holds the NextN layer, and the NextN tensors optional when the file only holds the trunk","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-12T21:04:41Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: drop (feature)","llm: publish; 3 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11474","title":"ggml-org/llama.cpp b11474","retrieved_at":"2026-10-12T21:02:06Z","sha256":"e3bdf2cb888aaa9621321e68c884710a2000904fd2357613f52cb9e7b90bbda6","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11474","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11474"},{"id":"rss:cohere-blog:666d6e7769a2","kind":"announcement","title":"In-Language Reasoning: Data Mixing Bridges | Cohere","date":"2026-10-06","observed_at":"2026-10-13T10:09:11Z","subject":{"type":"feed","ref":"cohere-blog","url":"https://cohere.com/blog/building-multilingual-bridges","entity":"cohere","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["open-weights"],"affects":["pytorch","transformers"],"summary":{"why":"Cohere ha liberado Tiny Aya L2-Thinker (3.35B), un modelo de razonamiento diseñado para 'pensar' en el idioma del usuario en 60 lenguas, evitando el sesgo de razonar siempre en inglés."},"highlights":[{"text":"Lanzamiento de Tiny Aya L2-Thinker (3.35B) con capacidad de razonamiento en el idioma del prompt en 60 lenguas.","quote":"Our resulting model, Tiny Aya L2-Thinker (3.35B), reasons in the prompt language over 93% of the time across 60 languages.","source":0},{"text":"Liberación de pesos del modelo y datos de razonamiento multilingüe.","quote":"We release the model weights and multilingual reasoning data to support further work on accessible, in-language reasoning.","source":0},{"text":"Alta eficiencia en idiomas de bajos recursos comparado con otros modelos.","quote":"Tiny Aya L2-Thinker is the only model that stays strong on all three axes as resource availability declines.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T10:09:11Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 3 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://cohere.com/blog/building-multilingual-bridges","title":"In-Language Reasoning: Data Mixing Bridges | Cohere","retrieved_at":"2026-10-13T09:59:45Z","sha256":"1ff9e75d1a716f0b96e5a71783907f52e0bed21ad51ebd1ebef1a2a38febc8f5","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--cohere-blog--666d6e7769a2","related":[],"org":{"slug":"cohere","name":"Cohere","kind":"laboratorio","url":"https://cohere.com"}},{"id":"rss:deepmind-models:52bf10e5b6de","kind":"announcement","title":"Nano Banana 2.1 - Model Card — Google DeepMind","date":"2026-10-06","observed_at":"2026-10-13T10:08:35Z","subject":{"type":"feed","ref":"deepmind-models","url":"https://deepmind.google/models/model-cards/nano-banana-2-1/","entity":"google","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":[],"affects":["gemini-cli"],"summary":{"why":"Se anuncia un nuevo modelo multimodal de la serie Gemini 3 (Nano Banana 2.1) con capacidades avanzadas de razonamiento, generación y edición de imágenes, disponible vía API."},"highlights":[{"text":"Modelo multimodal nativo de la serie Gemini 3 con capacidades de razonamiento.","quote":"Nano Banana 2.1 is a member of the Gemini 3 series of models, a suite of highly-capable, natively multimodal reasoning models.","source":0},{"text":"Soporta una ventana de contexto de hasta 1 millón de tokens para texto e imágenes.","quote":"Inputs Text strings (e.g., a prompt, document(s)) and images, with a token context window of up to 1M.","source":0},{"text":"Capacidad de salida tanto de texto como de imágenes (4K tokens para imagen).","quote":"Outputs Image, with a 4K token output and text, with a 64K token output.","source":0},{"text":"Distribución a través de API para desarrolladores.","quote":"Our models are available to downstream providers via an application program interface (API)","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T10:08:35Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://deepmind.google/models/model-cards/nano-banana-2-1/","title":"Nano Banana 2.1 - Model Card — Google DeepMind","retrieved_at":"2026-10-13T09:59:31Z","sha256":"1b126b310915225bbc6eaf50d879bb03fddd217b37f17945e167e825c80027a4","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--deepmind-models--52bf10e5b6de","related":[],"org":{"slug":"google","name":"Google","kind":"empresa","url":"https://deepmind.google"}},{"id":"rss:anthropic-news:2bd8707fdab8","kind":"announcement","title":"Expanding the Cyber Verification Program \\ Anthropic","date":"2026-10-06","observed_at":"2026-10-13T10:08:23Z","subject":{"type":"feed","ref":"anthropic-news","url":"https://www.anthropic.com/news/cyber-verification-program","entity":"anthropic","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":[],"affects":["claude-code"],"summary":{"why":"Anthropic lanza un programa de acceso con modelos menos restrictivos (Claude Opus 5.5, Sonnet 5.5 y Mythos 5.1) específicamente para profesionales de ciberseguridad, permitiendo tareas de red-teaming y análisis de vulnerabilidades que antes eran bloqueadas por sus filtros de seguridad."},"highlights":[{"text":"Nuevos niveles de acceso para profesionales de seguridad","quote":"The program now consists of three access tiers, which allow security teams to apply for the level of access that best suits their work.","source":0},{"text":"Modelos de vanguardia incluidos en el programa","quote":"Each tier includes access to our most capable models, including Claude Opus 5.5, Claude Sonnet 5.5, Claude Mythos 5.1, and new models moving forward.","source":0},{"text":"Reducción de bloqueos para tareas de defensa y red-teaming","quote":"In the Red Team Access tier, no blocks occurred, and Claude Opus 5.5 successfully completed 34 of the 50 tasks","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T10:08:23Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 3 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://www.anthropic.com/news/cyber-verification-program","title":"Expanding the Cyber Verification Program \\ Anthropic","retrieved_at":"2026-10-13T10:04:35Z","sha256":"38ba44fbea3da3cd849f258bccb517509116dbaea019c17177e007580ae21f82","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--anthropic-news--2bd8707fdab8","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"}},{"id":"rss:openai:00427e47df5e","kind":"announcement","title":"Sharing AI progress in mathematics","date":"2026-10-06","observed_at":"2026-10-13T09:11:05Z","subject":{"type":"feed","ref":"openai","url":"https://openai.com/index/sharing-ai-progress-in-mathematics","entity":"openai","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":[],"affects":[],"summary":{"why":"OpenAI comparte avances en resolución de problemas matemáticos mediante un modelo frontera y libera formalizaciones de pruebas en Lean, lo cual es relevante para la investigación en razonamiento lógico y verificación formal."},"highlights":[{"text":"Publicación de resultados sobre problemas abiertos en matemáticas mediante un modelo frontera interno.","quote":"OpenAI publishes new results on open problems in mathematics from an internal frontier model","source":0},{"text":"Liberación de formalizaciones de pruebas en Lean y detalles de investigación en GitHub.","quote":"shares Lean proof formalizations and research details on GitHub","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T09:11:05Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 2 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://openai.com/index/sharing-ai-progress-in-mathematics","title":"Sharing AI progress in mathematics","retrieved_at":"2026-10-13T09:10:19Z","sha256":"4db42ee12d42afe9dad2f82fddafaa7018c823e2a092ebc70d7157b664ceeb8d","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--openai--00427e47df5e","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"}},{"id":"rss:mistral:1ef243007176","kind":"announcement","title":"Introducing Mistral Large 4","date":"2026-10-06","observed_at":"2026-10-13T06:20:33Z","subject":{"type":"feed","ref":"mistral","url":"https://mistral.ai/news/mistral-large-4/","entity":"mistral","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":[],"summary":{"why":"El lanzamiento de un nuevo modelo de la serie Large de Mistral es un hito relevante para desarrolladores que buscan capacidades de razonamiento avanzadas en sus implementaciones."},"highlights":[{"text":"Lanzamiento de Mistral Large 4","quote":"Introducing Mistral Large 4","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:20:33Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://mistral.ai/news/mistral-large-4/","title":"Introducing Mistral Large 4","retrieved_at":"2026-10-13T05:51:54Z","sha256":"f5a96ed94535254f482f8bf32d36d0453db55943eb5fa3fe510c82145ddd30af","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--mistral--1ef243007176","related":[],"org":{"slug":"mistral","name":"Mistral AI","kind":"laboratorio","url":"https://mistral.ai"}},{"id":"github:ggml-org/llama.cpp:b11447","kind":"software_release","title":"llama.cpp b11447","date":"2026-10-06","observed_at":"2026-10-13T05:53:38Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11447","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"try","tags":["new-architecture"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":null},"highlights":[{"text":"Soporte para una nueva arquitectura de modelo","quote":"models: support pplx-decider (#30044)","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:53:38Z","rejected":[]},"triage":{"method":"rules","model":"gemma4","reasons":["reglas: publish (new-architecture)","llm lo descartó pero hay señal fuerte; se aplica la decisión de reglas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11447","title":"ggml-org/llama.cpp b11447","retrieved_at":"2026-10-13T05:51:12Z","sha256":"4254f0a3c5dfbe0a4a7ddec706020e983f6ec23066887e94ff2c295b9995d3c1","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11447","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11447"},{"id":"rss:deepmind:c11a83f73b42","kind":"announcement","title":"EmbeddingGemma 2: an open, lightweight multimodal embedding model","date":"2026-10-06","observed_at":"2026-10-12T21:09:11Z","subject":{"type":"feed","ref":"deepmind","url":"https://deepmind.google/blog/embeddinggemma-2-an-open-lightweight-multimodal-embedding-model/","entity":"google","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-architecture","new-model"],"affects":["pytorch","transformers"],"summary":{"why":"Google lanza un nuevo modelo de embeddings multimodal ligero y abierto, útil para mejorar la recuperación en sistemas RAG con capacidades multimodales."},"highlights":[{"text":"Lanzamiento de EmbeddingGemma 2, un modelo de embeddings multimodal abierto y ligero.","quote":"EmbeddingGemma 2: an open, lightweight multimodal embedding model","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-12T21:09:11Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas","publish con severidad minor: queda en observación","corrección editorial: lanzamiento real nunca es minor (la severidad minor era un fallo de calibración del prompt)"]},"sources":[{"url":"https://deepmind.google/blog/embeddinggemma-2-an-open-lightweight-multimodal-embedding-model/","title":"EmbeddingGemma 2: an open, lightweight multimodal embedding model","retrieved_at":"2026-10-12T21:02:35Z","sha256":"9af2ece9426382f4b0493c6375047cb70419b5f169d8a11a58d452410d3b78ec","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--deepmind--c11a83f73b42","related":["hf:google/embeddinggemma-2"],"org":{"slug":"google","name":"Google","kind":"empresa","url":"https://deepmind.google"}},{"id":"github:anthropics/claude-code:v2.1.291","kind":"software_release","title":"claude-code v2.1.291","date":"2026-10-06","observed_at":"2026-10-12T21:06:31Z","subject":{"type":"repo","ref":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.291","entity":"anthropic","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":[],"affects":["claude-code"],"summary":{"why":"Corrige errores críticos de regresión que causaban la pérdida de mensajes en las sesiones y fallos en los prompts de permisos."},"highlights":[{"text":"Se corrigió una regresión donde las sesiones en la nube podían omitir respuestas a prompts de permisos.","quote":"Fixed a regression in 2.1.290 where cloud sessions could drop answers to permission prompts","source":0},{"text":"Se corrigió una regresión donde se podían perder los últimos mensajes de una sesión al salir.","quote":"Fixed a regression in 2.1.288 where the last messages of a session could be lost when quitting","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-12T21:06:31Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: drop (sin señales)","llm: publish; 2 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.291","title":"anthropics/claude-code v2.1.291","retrieved_at":"2026-10-12T21:02:13Z","sha256":"de227a0c3298adbfa64cb132bef0153c90f33c1fe40da0cb582a42c71842423d","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--anthropics--claude-code--v2.1.291","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"},"tool":{"slug":"claude-code","name":"Claude Code"},"version":"v2.1.291"},{"id":"github:anthropics/claude-code:v2.1.292","kind":"software_release","title":"claude-code v2.1.292","date":"2026-10-06","observed_at":"2026-10-12T21:06:24Z","subject":{"type":"repo","ref":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.292","entity":"anthropic","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["feature","performance","security"],"affects":["claude-code"],"summary":{"why":"Esta actualización incluye parches de seguridad críticos sobre el bypass de permisos de lectura de archivos, mejoras significativas en el rendimiento de renderizado y nuevas capacidades de gestión de agentes y caché de prompts."},"highlights":[{"text":"Corrección de seguridad en la aprobación de hooks PreToolUse y el bypass del modo automático para lecturas de archivos desde rutas de red (UNC).","quote":"Security: Fixed PreToolUse hook approvals and auto mode bypassing the permission prompt for file reads from network (UNC) paths","source":0},{"text":"Introducción de soporte para prompt caching en mods para optimizar peticiones.","quote":"Added prompt caching to `$.model.complete` for mods: `prompt` and `system` take blocks of text, and `cache: true` on a block caches the request up to it","source":0},{"text":"Mejora en la velocidad de renderizado de respuestas largas con listas.","quote":"Improved rendering speed of long bulleted or numbered replies: they stream, resize and re-open in the transcript (ctrl+o) much faster","source":0},{"text":"Nuevo parámetro de esfuerzo para el Agent tool para controlar la ejecución de sub-agentes.","quote":"Added an `effort` parameter to the Agent tool, so Claude runs a sub-agent at the effort level you ask for","source":0},{"text":"Corrección de vulnerabilidad en el sandbox que permitía leer archivos subidos bajo rutas específicas.","quote":"Fixed subagent definitions with `permissionMode: auto` entering auto mode when auto mode is unavailable","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-12T21:06:24Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (feature)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.292","title":"anthropics/claude-code v2.1.292","retrieved_at":"2026-10-12T21:02:13Z","sha256":"b43c92233f08b4d1025824dcdc6896dc2cf02f8043d7fd3501a0a7f398c9d610","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--anthropics--claude-code--v2.1.292","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"},"tool":{"slug":"claude-code","name":"Claude Code"},"version":"v2.1.292"},{"id":"github:huggingface/transformers:v5.19.0","kind":"software_release","title":"transformers Release v5.19.0","date":"2026-10-06","observed_at":"2026-10-12T21:05:25Z","subject":{"type":"repo","ref":"huggingface/transformers","url":"https://github.com/huggingface/transformers/releases/tag/v5.19.0","entity":"huggingface","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["breaking","performance"],"affects":["transformers"],"summary":{"why":"Esta versión introduce cambios disruptivos en el manejo de logits para modelos MoE y en la implementación de atención (depreciación de prefijos 'paged|'), además de mejoras significativas en el soporte de continuous batching y paralelismo de expertos."},"highlights":[{"text":"Cambio en la salida de logits para todos los modelos MoE que los computan.","quote":"All MoE models whose routers compute logits now return them when `output_router_logits=True`, following the Qwen3-MoE pattern","source":0},{"text":"Depreciación del prefijo 'paged|' para implementaciones de SDPA y Flash Attention.","quote":"The `\"paged|\"` prefix for SDPA and flash attention implementations is deprecated, so users should set the regular attention implementation (e.g. `sdpa` or `flash_attention_2`) for continuous batching","source":0},{"text":"Soporte directo para continuous batching en funciones de atención regulares.","quote":"The regular flash and SDPA attention functions (`flash_attention.py`, `sdpa_attention.py`) now support continuous batching directly","source":0},{"text":"Nuevas capacidades de paralelismo de expertos (Expert Parallelism) con token-dispatch.","quote":"Expert parallelism gains a token-dispatch implementation, selected via the new `ep_dispatch_experts` plan rule","source":0},{"text":"Adición del modelo multimodal EmbeddingGemma 2.","quote":"EmbeddingGemma 2 is a multimodal embedding model from Google built on the Gemma 4 architecture.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-12T21:05:25Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, deprecation)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/huggingface/transformers/releases/tag/v5.19.0","title":"huggingface/transformers Release v5.19.0","retrieved_at":"2026-10-12T21:02:11Z","sha256":"1bf92cd2f333fc3f4d683b416d918365fdf6105681237a947b8e687ffab9eada","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--huggingface--transformers--v5.19.0","related":["hf:google/embeddinggemma-2"],"org":{"slug":"huggingface","name":"Hugging Face","kind":"empresa","url":"https://huggingface.co"},"tool":{"slug":"transformers","name":"Transformers"},"version":"v5.19.0"},{"id":"github:ggml-org/llama.cpp:b11454","kind":"software_release","title":"llama.cpp b11454","date":"2026-10-06","observed_at":"2026-10-12T21:04:45Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11454","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"update","tags":["fix","new-model"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Añade soporte completo para la arquitectura K2 Horizon (incluyendo conversión GGUF y soporte de herramientas) y corrige un error crítico de regex que impedía la carga de modelos en Windows."},"highlights":[{"text":"Soporte para modelos K2 Horizon dense y MoVA","quote":"model : add K2 Horizon dense and MoVA support","source":0},{"text":"Corrección de error de carga en Windows (MSVC)","quote":"On MSVC std::regex rejects \\p{...}, so no K2-Horizon GGUF loads on Windows at all","source":0},{"text":"Soporte para razonamiento y llamadas a herramientas en K2 Horizon","quote":"chat : support K2 Horizon reasoning and tool calls","source":0},{"text":"Mejoras en el splitter de unicode para K2-Horizon","quote":"unicode : add the K2-Horizon pre-tokenizer splitter","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-12T21:04:45Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (new-architecture)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11454","title":"ggml-org/llama.cpp b11454","retrieved_at":"2026-10-12T21:02:06Z","sha256":"2876d6ec1da82958f79aa78ad9a38e5e91876d07af32ae9dc24b2bfb96f2415a","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11454","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11454"},{"id":"rss:cohere-blog:0a07e9cff5f9","kind":"announcement","title":"North 2: Enterprise AI Without Compromises","date":"2026-10-05","observed_at":"2026-10-13T10:09:16Z","subject":{"type":"feed","ref":"cohere-blog","url":"https://cohere.com/blog/introducing-north-2","entity":"cohere","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":[],"affects":[],"summary":{"why":"Cohere lanza North 2, una plataforma de agentes que permite despliegues on-premises, en la nube o air-gapped, siendo agnóstica al modelo (permite usar modelos propios) y optimizada para hardware NVIDIA."},"highlights":[{"text":"Plataforma agnóstica al modelo que permite traer modelos propios.","quote":"Model agnostic: Whether you’re using Cohere’s models — or bringing your own — use the models that support your team’s best work.","source":0},{"text":"Soporte para despliegues locales, en VPC o entornos aislados (air-gapped).","quote":"You’ll get full control over where and how AI runs, with flexible cloud, on‑premises, and air‑gapped deployments.","source":0},{"text":"Nuevo sistema de orquestación para agentes multi-paso.","quote":"North 2 has a redesigned orchestration system enabling more sophisticated, reliable, multi-step automations and agents across the enterprise.","source":0},{"text":"Optimización de rendimiento y tokens en hardware NVIDIA.","quote":"When using NVIDIA hardware with Cohere models, get increased tokens per second per node, effectively decreasing spend.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T10:09:16Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://cohere.com/blog/introducing-north-2","title":"North 2: Enterprise AI Without Compromises","retrieved_at":"2026-10-13T09:59:45Z","sha256":"dc8cf706783c1f5d01ad916d64924a4255b36e1eee87c842ad4b660b178879b0","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--cohere-blog--0a07e9cff5f9","related":[],"org":{"slug":"cohere","name":"Cohere","kind":"laboratorio","url":"https://cohere.com"}},{"id":"hf:inclusionAI/LemonDecider","kind":"model_release","title":"inclusionAI publica LemonDecider","date":"2026-10-05","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"inclusionAI/LemonDecider","url":"https://huggingface.co/inclusionAI/LemonDecider","entity":"inclusionai","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 4.5 B parámetros, licencia apache-2.0, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":4539918336,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"apache-2.0","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"feature-extraction","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-10-05T04:27:12.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"AutoJevForDecisionClassification","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/inclusionAI/LemonDecider","title":"Hugging Face · inclusionAI/LemonDecider","retrieved_at":"2026-10-13T08:55:15Z","sha256":"1cfcec0d9781ebc884611c2b5a4d7adb60854a4b0ed4422f052da3e11f8d1ad3","license":null},{"url":"https://huggingface.co/inclusionAI/LemonDecider/blob/main/config.json","title":"inclusionAI/LemonDecider · config.json","retrieved_at":"2026-10-13T08:55:15Z","sha256":"21fb2a3497338b800c1d080ab2d0157f2658f2443ee333e21cae37fc4acda6a2","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"inclusionAI/LemonDecider","params_total":4539918336,"architecture":"AutoJevForDecisionClassification","context_length":262144,"license":"apache-2.0","gated":false,"pipeline":"feature-extraction","variants":[],"kv_bytes_per_token":32768,"quants":[]},"slug":"hf--inclusionai--lemondecider","related":[],"org":{"slug":"inclusionai","name":"inclusionAI (Ant Group)","kind":"laboratorio","url":"https://github.com/inclusionAI"}},{"id":"github:vllm-project/vllm:v0.31.0","kind":"software_release","title":"vllm v0.31.0","date":"2026-10-05","observed_at":"2026-10-13T05:56:18Z","subject":{"type":"repo","ref":"vllm-project/vllm","url":"https://github.com/vllm-project/vllm/releases/tag/v0.31.0","entity":"vllm","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["breaking","new-model","performance"],"affects":["vllm"],"summary":{"why":"Es una actualización mayor de vLLM con optimizaciones críticas de rendimiento para modelos DeepSeek-V4.1 y soporte para nuevos arquitecturas, además de cambios que rompen la compatibilidad en la configuración de cuantización y parámetros multimodales."},"highlights":[{"text":"Optimización masiva para DeepSeek-V4.1-Flash mediante FlashMLA y caché KV comprimido.","quote":"FlashMLA mega attention with the V4.1 NVFP4 compressed KV cache is now the SM100 default","source":0},{"text":"Nuevo CLI 'vllm preload' para mantener pesos en GPU y acelerar reinicios.","quote":"the new `vllm preload` CLI launches the weight-cache daemon that keeps post-quantized weights resident in GPU memory across engine restarts","source":0},{"text":"Introducción de Model Runner V2 con soporte para speculative decoding.","quote":"Model Runner V2 and speculative decoding: draft-model speculative decoding (#43091) and custom logits processors (#56497) on Model Runner V2","source":0},{"text":"Cambios que rompen la compatibilidad (Breaking changes) en parámetros multimodales y cuantización.","quote":"Breaking changes: per-request multimodal kwargs gated (#58830); `tokenizer_mode=\"slow\"` removed (#58545)","source":0},{"text":"Soporte para nuevos modelos como Gemma4, Kimi-K3 y DiffusionGemma.","quote":"New model capabilities: DiffusionGemma structured generation mode with bounded single-token choices (#57250)","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:56:18Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, performance)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/vllm-project/vllm/releases/tag/v0.31.0","title":"vllm-project/vllm v0.31.0","retrieved_at":"2026-10-13T05:51:17Z","sha256":"63bbe33b88ffba6dd42d7529517fb38b00f07ad124d9d70274544c4b94ba0708","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--vllm-project--vllm--v0.31.0","related":[],"org":{"slug":"vllm","name":"vLLM","kind":"proyecto","url":"https://vllm.ai"},"tool":{"slug":"vllm","name":"vLLM"},"version":"v0.31.0"},{"id":"github:ggml-org/llama.cpp:b11405","kind":"software_release","title":"llama.cpp b11405","date":"2026-10-05","observed_at":"2026-10-13T05:54:11Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11405","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"try","tags":["hardware","performance"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Esta actualización incluye optimizaciones críticas en los kernels de CUDA y ROCm (Lightning indexer), mejorando significativamente la velocidad en GPUs AMD y optimizando el uso de registros en CUDA."},"highlights":[{"text":"Optimización del indexador Lightning para CUDA con soporte de precisión flotante para evitar desbordamientos.","quote":"the half2 products overflow once a single q * k exceeds the f16 range. The queries stay in float in shared memory and each half2 of keys is widened once for both tokens, so every product and sum is computed in float.","source":0},{"text":"Mejora masiva de rendimiento en GPUs AMD (ROCm) al evitar el desbordamiento de registros.","quote":"makes the kernel 36x faster on an R9700 and slightly faster on CUDA.","source":0},{"text":"Optimización de memoria y velocidad mediante el uso de un solo barrier para todas las cabezas.","quote":"The tile kernel stages the queries and weights of every head at once, so each key element is widened from half once and feeds all heads, with a single barrier.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:54:11Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: drop (performance)","llm: publish; 3 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11405","title":"ggml-org/llama.cpp b11405","retrieved_at":"2026-10-13T05:51:12Z","sha256":"d2f53523eac830a8684178a29aa6370845a21dc46e4641c7728d6a65af284e01","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11405","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11405"},{"id":"github:ggml-org/llama.cpp:v0.6.0","kind":"software_release","title":"llama.cpp v0.6.0","date":"2026-10-05","observed_at":"2026-10-13T05:53:42Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/v0.6.0","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"update","tags":["api","new-model","performance"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Esta versión introduce cambios estructurales en la API de batches, soporte para modelos híbridos masivos (GLM-5.3-Flash 320B) y optimizaciones críticas de rendimiento en Metal y Vulkan."},"highlights":[{"text":"Nueva API de batches extendida para inputs mixtos de tokens y embeddings.","quote":"New `llama_batch_ext` extended batch API (with `llama_process`) for mixed token/embedding inputs","source":0},{"text":"Soporte para el modelo híbrido GLM-5.3-Flash de 320B parámetros.","quote":"adds support for the GLM-5.3-Flash (GLM5-Next) 320B hybrid model","source":0},{"text":"Mejoras de rendimiento de hasta 3x en mat-mul para Apple GPUs mediante nuevos kernels MMA.","quote":"new few-row MMA mat-mul kernels for speculative and batched decoding, up to ~3x faster mat-mul on Apple GPUs","source":0},{"text":"Nuevo endpoint `/v1/systemone` para modelos de decisión.","quote":"ships a new `/v1/systemone` server API for decision models","source":0},{"text":"Soporte de Flash Attention disperso para K/V cuantizado en Vulkan.","quote":"sparse flash attention for quantized K/V on Vulkan","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:53:42Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (deprecation, feature, new-architecture, performance)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/v0.6.0","title":"ggml-org/llama.cpp v0.6.0","retrieved_at":"2026-10-13T05:51:12Z","sha256":"12400f7c722cbb0f4eaaa0acfb22104475356785223844a81be77b3c22aebe0d","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--v0.6.0","related":["hf:zai-org/GLM-5.3","hf:zai-org/GLM-5.3-Flash"],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"v0.6.0"},{"id":"github:anthropics/claude-code:v2.1.290","kind":"software_release","title":"claude-code v2.1.290","date":"2026-10-05","observed_at":"2026-10-12T21:06:36Z","subject":{"type":"repo","ref":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.290","entity":"anthropic","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["fix","performance","security"],"affects":["claude-code"],"summary":{"why":"Esta versión corrige vulnerabilidades críticas de seguridad relacionadas con la lectura de archivos fuera del directorio de trabajo (sandbox escape) y mejora la estabilidad en sesiones largas y el manejo de proxies."},"highlights":[{"text":"Corrección de fuga de seguridad en lectura de archivos mediante enlaces.","quote":"Fixed an image read on macOS and Windows being able to return a file outside what was approved, through a link swapped in mid-read","source":0},{"text":"Corrección de bypass de permisos en menciones con @.","quote":"Fixed an `@`-mention under the read block or `--restricted` being able to read a file outside the working directories through a link changed mid-read","source":0},{"text":"Mejora en la estabilidad de sesiones largas con imágenes.","quote":"Fixed long sessions with hundreds of images getting stuck on \"Request rejected as unprocessable by the model\" errors","source":0},{"text":"Corrección de fallos de conexión tras proxies o gateways.","quote":"Fixed requests failing behind proxies and gateways that reject one of Claude Code's beta headers","source":0},{"text":"Corrección de errores de permisos en comandos de Bash.","quote":"Bash permission checks auto-approving some read-only commands (such as `rg` or `git grep`) whose arguments the shell would still expand as wildcards; these now prompt for approval","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-12T21:06:36Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (feature)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.290","title":"anthropics/claude-code v2.1.290","retrieved_at":"2026-10-12T21:02:13Z","sha256":"d8b98ae66b0c0adf5a5d29bd74164e5a99ddc6de761332756cde43fa599e217e","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--anthropics--claude-code--v2.1.290","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"},"tool":{"slug":"claude-code","name":"Claude Code"},"version":"v2.1.290"},{"id":"github:ggml-org/llama.cpp:b11364","kind":"software_release","title":"llama.cpp b11364","date":"2026-10-03","observed_at":"2026-10-13T05:55:09Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11364","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"try","tags":["new-architecture"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":null},"highlights":[{"text":"Soporte para una nueva arquitectura de modelo","quote":"model: support nimble decision model (#29844)","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:55:09Z","rejected":[]},"triage":{"method":"rules","model":"gemma4","reasons":["reglas: publish (new-architecture)","llm lo descartó pero hay señal fuerte; se aplica la decisión de reglas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11364","title":"ggml-org/llama.cpp b11364","retrieved_at":"2026-10-13T05:51:13Z","sha256":"7e07fd905fdbdf0998ed77d0d18538ae74091d0cb40471f533918ebb2aae700d","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11364","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11364"},{"id":"github:ggml-org/llama.cpp:b11371","kind":"software_release","title":"llama.cpp b11371","date":"2026-10-03","observed_at":"2026-10-13T05:54:56Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11371","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"try","tags":["new-architecture"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":null},"highlights":[{"text":"Soporte para una nueva arquitectura de modelo","quote":"model: add support for clef decision model (text-only) (#29831)","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:54:56Z","rejected":[]},"triage":{"method":"rules","model":"gemma4","reasons":["reglas: publish (new-architecture)","llm lo descartó pero hay señal fuerte; se aplica la decisión de reglas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11371","title":"ggml-org/llama.cpp b11371","retrieved_at":"2026-10-13T05:51:13Z","sha256":"68ee2db029faba88b5a011c4ac7bf3580a6de438aa6b07b45cf1806a0addadb3","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11371","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11371"},{"id":"rss:deepmind-models:744e7330e24d","kind":"announcement","title":"Gemini 3.8 Audio - Model Card — Google DeepMind","date":"2026-10-02","observed_at":"2026-10-13T10:08:36Z","subject":{"type":"feed","ref":"deepmind-models","url":"https://deepmind.google/models/model-cards/gemini-3-8-audio/","entity":"google","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":[],"affects":["gemini-cli"],"summary":{"why":"Se anuncia una nueva generación de modelos de la familia Gemini (3.8) con capacidades nativas de audio, video y salida de voz (TTS), lo que cambia las opciones de evaluación para desarrolladores de aplicaciones multimodales."},"highlights":[{"text":"Lanzamiento de la familia Gemini 3.8 Audio con capacidades multimodales nativas.","quote":"Gemini 3.8 Audio (Live, Live Extended Thinking, Flash TTS, Flash-Lite TTS) is an addition to the Gemini 3 series of highly-capable models.","source":0},{"text":"Soporte para entrada de audio, imagen, video y texto con ventana de contexto de 128K.","quote":"Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking: Audio, images, video, and text with a token context window of up to 128K.","source":0},{"text":"Capacidad de salida de video con avatar para interacciones naturales.","quote":"With Live Avatar: Audio, video, and text, with 24K token output.","source":0},{"text":"Disponibilidad a través de API para integración en aplicaciones.","quote":"Our models are available to downstream providers via an application program interface (API) and subject to relevant terms of use.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T10:08:36Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://deepmind.google/models/model-cards/gemini-3-8-audio/","title":"Gemini 3.8 Audio - Model Card — Google DeepMind","retrieved_at":"2026-10-13T09:59:31Z","sha256":"3d377d7214b073c43677961753a02d121843dffcde7b33bac2faa2e7b89b2854","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--deepmind-models--744e7330e24d","related":[],"org":{"slug":"google","name":"Google","kind":"empresa","url":"https://deepmind.google"}},{"id":"rss:openai:c38af6e5cae4","kind":"announcement","title":"A model guide for the GPT-6 family","date":"2026-10-02","observed_at":"2026-10-13T09:11:17Z","subject":{"type":"feed","ref":"openai","url":"https://openai.com/index/practical-guide-building-gpt-6","entity":"openai","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":[],"affects":[],"summary":{"why":"El anuncio introduce una nueva familia de modelos (GPT-6) y proporciona directrices para la producción, el ajuste de razonamiento y la coordinación de herramientas, lo que impacta en el diseño de workflows de IA."},"highlights":[{"text":"Guía sobre cómo elegir modelos GPT-6, ajustar el esfuerzo de razonamiento y mejorar prompts.","quote":"Learn how startups can choose GPT-6 models, tune reasoning effort, improve prompts and skills, coordinate tools, and prepare workflows for production.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T09:11:17Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://openai.com/index/practical-guide-building-gpt-6","title":"A model guide for the GPT-6 family","retrieved_at":"2026-10-13T09:10:19Z","sha256":"9649b7a7c88b8a6ef6f2ed1fbb6d1e770e3c8c2e07f606a3b6fa402f346eb47d","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--openai--c38af6e5cae4","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"}},{"id":"rss:pytorch:bd5a46541bfe","kind":"announcement","title":"Building a High-Performance and Portable vLLM Linear Backend with Helion","date":"2026-10-02","observed_at":"2026-10-13T06:21:19Z","subject":{"type":"feed","ref":"pytorch","url":"https://pytorch.org/blog/building-a-high-performance-and-portable-vllm-linear-backend-with-helion/","entity":"pytorch","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-architecture","performance"],"affects":["pytorch","vllm"],"summary":{"why":"La integración de Helion en el backend de vLLM busca mejorar el rendimiento de la inferencia mediante un DSL de kernels autotunizado, lo que podría optimizar la ejecución de modelos en este motor."},"highlights":[{"text":"Integración de Helion en el backend lineal de vLLM para mejorar el rendimiento de la inferencia.","quote":"We integrated Helion into vLLM’s linear backend to explore how an autotuned, high-level kernel DSL can improve LLM inference performance","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:21:19Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas","publish con severidad minor: queda en observación","corrección editorial: lanzamiento real nunca es minor (la severidad minor era un fallo de calibración del prompt)"]},"sources":[{"url":"https://pytorch.org/blog/building-a-high-performance-and-portable-vllm-linear-backend-with-helion/","title":"Building a High-Performance and Portable vLLM Linear Backend with Helion","retrieved_at":"2026-10-13T05:51:55Z","sha256":"5229d31c724bfe82518b3327448c5ef6254c792cd71d1b7d996c4b575233384f","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--pytorch--bd5a46541bfe","related":[],"org":{"slug":"pytorch","name":"PyTorch","kind":"proyecto","url":"https://pytorch.org"}},{"id":"rss:allenai:d185427efe0f","kind":"announcement","title":"Open-sourcing AstaBrief, the fast report-generation model in Asta","date":"2026-10-02","observed_at":"2026-10-13T06:20:55Z","subject":{"type":"feed","ref":"allenai","url":"https://allenai.org/blog/astabrief","entity":"allenai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model","open-weights"],"affects":["pytorch","transformers"],"summary":{"why":"Es un modelo de 8B parámetros especializado en la generación de informes científicos con citas, disponible para ejecución local."},"highlights":[{"text":"Lanzamiento de AstaBrief, un modelo de 8B parámetros con pesos abiertos para informes científicos.","quote":"We’re releasing AstaBrief, an 8B open-weights model for generating cited scientific reports","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:20:55Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas","publish con severidad minor: queda en observación","corrección editorial: lanzamiento real nunca es minor (la severidad minor era un fallo de calibración del prompt)","tag open-weights añadido: la cita verificada declara pesos abiertos"]},"sources":[{"url":"https://allenai.org/blog/astabrief","title":"Open-sourcing AstaBrief, the fast report-generation model in Asta","retrieved_at":"2026-10-13T05:51:55Z","sha256":"58917ffae2774b085f13244819a1ddae53410a5da6e6274b40cacff1be21eb2f","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--allenai--d185427efe0f","related":[],"org":{"slug":"allenai","name":"Allen Institute for AI","kind":"laboratorio","url":"https://allenai.org"}},{"id":"github:sgl-project/sglang:v0.5.21","kind":"software_release","title":"sglang v0.5.21","date":"2026-10-02","observed_at":"2026-10-13T06:12:39Z","subject":{"type":"repo","ref":"sgl-project/sglang","url":"https://github.com/sgl-project/sglang/releases/tag/v0.5.21","entity":"sglang","channel":"stable"},"status":"published","severity":"notable","action":"try","tags":["performance","security"],"affects":["sglang"],"summary":{"why":"Esta versión introduce mejoras críticas de rendimiento en modelos DeepSeek y Kimi, soporte para nuevos modelos (incluyendo modelos de difusión), y una nueva API de decisiones para clasificación de baja latencia."},"highlights":[{"text":"Nuevos modelos soportados (LLM, VLM y Diffusion)","quote":"DeepSeek-V4.1 Flash | LLM / VLM","source":0},{"text":"Mejora de rendimiento en DeepSeek-V4.1","quote":"DeepSeek-V4.1 gets 22% faster first token on long prompts","source":0},{"text":"Nueva API de Decisiones para clasificación de baja latencia","quote":"New Decisions API (`/v1/decisions`) turns an LLM / VLM into a low-latency classifier and scorer","source":0},{"text":"Optimización de throughput para Kimi K3","quote":"Kimi K3 gets 20.6% higher prefill throughput in PD serving","source":0},{"text":"Soporte para hardware AMD MI355X con FP8/MXFP4","quote":"GLM-5.3-Flash now runs on AMD MI355X, with FP8 / MXFP4 MoE and MTP speculative decoding","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:12:39Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, deprecation, feature, new-architecture, performance)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/sgl-project/sglang/releases/tag/v0.5.21","title":"sgl-project/sglang v0.5.21","retrieved_at":"2026-10-13T05:51:38Z","sha256":"cc9a2e43f63f29e81795fd67108bafbf603cc5f22c7d63d06c66787d0dc861a8","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--sgl-project--sglang--v0.5.21","related":["hf:deepseek-ai/DeepSeek-V4.1-Flash","hf:zai-org/GLM-5.3","hf:zai-org/GLM-5.3-Flash"],"org":{"slug":"sglang","name":"SGLang","kind":"proyecto","url":"https://sglang.ai"},"tool":{"slug":"sglang","name":"SGLang"},"version":"v0.5.21"},{"id":"github:openvinotoolkit/openvino:2026.4.1","kind":"software_release","title":"openvino 2026.4.1","date":"2026-10-01","observed_at":"2026-10-13T06:14:31Z","subject":{"type":"repo","ref":"openvinotoolkit/openvino","url":"https://github.com/openvinotoolkit/openvino/releases/tag/2026.4.1","entity":"intel","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":[],"affects":["openvino"],"summary":{"why":"Esta versión corrige un error crítico de corrupción de memoria/crash en GPUs y un fallo de estabilidad en NPUs de Intel Core Ultra. Además, mejora el soporte para modelos LLM modernos mediante optimizaciones en GroupQueryAttention y cuantización."},"highlights":[{"text":"Corrección de error crítico en el KV-cache que causaba corrupción de memoria o crashes en GPU.","quote":"an out-of-bounds `ScatterUpdate` index that could cause memory corruption or a driver crash on GPUs","source":0},{"text":"Solución de crashes inesperados en procesadores Intel Core Ultra Series 3 con drivers de NPU específicos.","quote":"Fixed unexpected application crashes on Intel® Core™ Ultra Series 3 processors on Windows that could occur with certain Intel® NPU PV drivers","source":0},{"text":"Mejoras en el soporte de GroupQueryAttention para modelos LLM a través de ONNX.","quote":"Improved `GroupQueryAttention` support for large language models served through the ONNX frontend","source":0},{"text":"Ampliación del soporte para modelos cuantizados mediante la extensión de MatMulNBits con bfloat16.","quote":"Extended `MatMulNBits` with `bfloat16` support","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:14:31Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (feature)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/openvinotoolkit/openvino/releases/tag/2026.4.1","title":"openvinotoolkit/openvino 2026.4.1","retrieved_at":"2026-10-13T05:51:46Z","sha256":"a1a39704e49d0594ee0494e31fd27074aad88cb841b1c1ab7747eead501b58f2","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--openvinotoolkit--openvino--2026.4.1","related":[],"org":{"slug":"intel","name":"Intel","kind":"empresa","url":"https://www.intel.com"},"tool":{"slug":"openvino","name":"OpenVINO"},"version":"2026.4.1"},{"id":"github:anthropics/claude-code:v2.1.287","kind":"software_release","title":"claude-code v2.1.287","date":"2026-10-01","observed_at":"2026-10-13T05:57:01Z","subject":{"type":"repo","ref":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.287","entity":"anthropic","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["performance","security"],"affects":["claude-code"],"summary":{"why":"Esta actualización introduce soporte para plugins (Claude Mods), mejora la gestión de memoria para archivos grandes y corrige una vulnerabilidad de seguridad crítica en el comando `rm` de Bash."},"highlights":[{"text":"Introducción de Claude Mods para permitir que los plugins modifiquen comportamientos profundos.","quote":"Added Claude Mods: plugins may now modify deeper behavior","source":0},{"text":"Corrección de seguridad en el comando `rm` que perdía su salvaguarda de confirmación en ciertos casos.","quote":"Fixed a dangerous `rm` (such as one on `/` or the home directory) losing its always-ask safeguard when the same command also redirected output to a `~` or wildcard path","source":0},{"text":"Mejora en el rendimiento de archivos grandes mediante streaming desde disco en lugar de carga en memoria.","quote":"large files now stream from disk instead of being read into memory","source":0},{"text":"Cambio en el contexto por defecto para modelos Opus 4.7+ y Fable a 1M de tokens en proveedores específicos.","quote":"Changed Opus 4.7+ and Fable to use a 1M context window by default on Bedrock, Vertex, Foundry and the Claude apps gateway","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:57:01Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (feature)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.287","title":"anthropics/claude-code v2.1.287","retrieved_at":"2026-10-13T05:51:19Z","sha256":"653eea440af88f458ea76a83ef311a049448fdc8ad52bfa2a3627cddb1127cd1","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--anthropics--claude-code--v2.1.287","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"},"tool":{"slug":"claude-code","name":"Claude Code"},"version":"v2.1.287"},{"id":"rss:cohere-blog:fa4d4be6d421","kind":"announcement","title":"RCP-nDCG@10: A New Approach to Enterprise Retrieval Quality | Cohere","date":"2026-09-30","observed_at":"2026-10-13T10:09:22Z","subject":{"type":"feed","ref":"cohere-blog","url":"https://cohere.com/blog/rcp-ndcg","entity":"cohere","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":[],"affects":[],"summary":{"why":"Cohere introduce una nueva metodología de evaluación (RCP-nDCG@10) para sistemas de recuperación (RAG/Search) que utiliza un juez de IA calibrado. Esto cambia la forma en que se mide la calidad de los modelos de embedding y reranking, permitiendo identificar documentos relevantes que los benchmarks tradicionales omiten."},"highlights":[{"text":"Introducción de RCP-nDCG@10 para medir la calidad de recuperación empresarial mediante un juez de IA calibrado.","quote":"we are introducing Rubric-Calibrated Preferences nDCG@10 (RCP-nDCG@10), an evaluation methodology that grades each retrieved document against the same explicit relevance criteria for every query, using a calibrated AI judge.","source":0},{"text":"La métrica resuelve el problema de los 'labels' incompletos en benchmarks tradicionales.","quote":"RCP-nDCG@10 addresses this coverage gap while preserving the ordering objective of traditional nDCG.","source":0},{"text":"Validación superior frente al juicio humano en comparación con el nDCG estándar.","quote":"Where RCP-nDCG and traditional nDCG named different winners, reviewers sided with RCP-nDCG 70% of the time.","source":0},{"text":"El código y los datos de la metodología están disponibles en GitHub.","quote":"Code and data: https://github.com/cohere-ai/rcp-ndcg","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T10:09:22Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://cohere.com/blog/rcp-ndcg","title":"RCP-nDCG@10: A New Approach to Enterprise Retrieval Quality | Cohere","retrieved_at":"2026-10-13T09:59:46Z","sha256":"f5db27ca08e5afa19d99baf224cf356f820aedd6a4f8b9045f0d52945f24ebbd","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--cohere-blog--fa4d4be6d421","related":[],"org":{"slug":"cohere","name":"Cohere","kind":"laboratorio","url":"https://cohere.com"}},{"id":"rss:deepmind:ce56aa83d006","kind":"announcement","title":"Gemini 4 Argon: our next era of frontier intelligence","date":"2026-09-30","observed_at":"2026-10-13T09:22:01Z","subject":{"type":"feed","ref":"deepmind","url":"https://deepmind.google/blog/gemini-4-argon-our-next-era-of-frontier-intelligence/","entity":"google","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":[],"affects":[],"summary":{"why":"Se anuncia la nueva generación de modelos frontera de Google (Gemini 4), lo que implica un cambio en el estado del arte que los desarrolladores deben evaluar para sus implementaciones de IA."},"highlights":[{"text":"Anuncio de la nueva era de inteligencia frontera con Gemini 4 Argon.","quote":"Gemini 4 Argon: our next era of frontier intelligence","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T09:22:01Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://deepmind.google/blog/gemini-4-argon-our-next-era-of-frontier-intelligence/","title":"Gemini 4 Argon: our next era of frontier intelligence","retrieved_at":"2026-10-13T09:19:34Z","sha256":"bfa4897eaff27ead9c9a2b51048bec2d3255b32716750147d933a6ed24613f87","license":null},{"url":"https://deepmind.google/blog/gemini-4-argon-our-next-era-of-frontier-intelligence/","title":"Introducing Gemini 4 Argon","retrieved_at":"2026-10-13T09:19:35Z","sha256":"51525685972b8de8984427dd8e61b4d6aaa2088a29031b4ce991a03ac196a7e6","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--deepmind--ce56aa83d006","related":[],"org":{"slug":"google","name":"Google","kind":"empresa","url":"https://deepmind.google"}},{"id":"rss:cohere-changelog:761d3b01fb63","kind":"announcement","title":"Announcing Cohere's Embed 5 Models | Cohere","date":"2026-09-30","observed_at":"2026-10-13T06:23:40Z","subject":{"type":"feed","ref":"cohere-changelog","url":"https://docs.cohere.com/changelog/embed-v5","entity":"cohere","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":[],"summary":{"why":"Cohere lanza su nueva generación de modelos de embeddings (Embed 5), lo cual es relevante para desarrolladores que implementan sistemas RAG o búsqueda semántica."},"highlights":[{"text":"Lanzamiento de los nuevos modelos Embed 5 Pro y Embed 5 Fast.","quote":"Release announcement for Embed 5 Pro and Embed 5 Fast, Cohere","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:23:40Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas","publish con severidad minor: queda en observación","corrección editorial: lanzamiento real nunca es minor (la severidad minor era un fallo de calibración del prompt)"]},"sources":[{"url":"https://docs.cohere.com/changelog/embed-v5","title":"Announcing Cohere's Embed 5 Models | Cohere","retrieved_at":"2026-10-13T05:52:20Z","sha256":"cd2836900f3bc4b03a8423540d9b74fefa6cc1152545656d852ea861b35fdcb2","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--cohere-changelog--761d3b01fb63","related":[],"org":{"slug":"cohere","name":"Cohere","kind":"laboratorio","url":"https://cohere.com"}},{"id":"github:huggingface/transformers:v5.18.0","kind":"software_release","title":"transformers Release 5.18.0","date":"2026-09-30","observed_at":"2026-10-13T05:56:35Z","subject":{"type":"repo","ref":"huggingface/transformers","url":"https://github.com/huggingface/transformers/releases/tag/v5.18.0","entity":"huggingface","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["breaking","feature","new-model"],"affects":["transformers"],"summary":{"why":"Esta versión de Transformers introduce soporte para nuevos modelos multimodales y de diarización de audio, además de incluir cambios disruptivos (breaking changes) en el procesamiento de imágenes y soporte para ROCm/vLLM que requieren atención."},"highlights":[{"text":"Soporte para el nuevo modelo de diarización de audio Nemotron 3.","quote":"Nemotron 3 Diarization is an open-weight streaming speaker diarization model designed to determine \"who spoke when\" in real-world audio.","source":0},{"text":"Integración de HyperCLOVAX Vision V2 para razonamiento multimodal.","quote":"HyperCLOVAX Vision V2 is a multimodal vision-language model developed by NAVER.","source":0},{"text":"Soporte para MoE (Mixture of Experts) en la integración de GGUF.","quote":"Support for MoE in the GGUF integration (#48529)","source":0}],"facts":[],"verification":{"status":"partial","checked_at":"2026-10-13T05:56:35Z","rejected":["NemotronH Omni is a multimodal reasoning model from NVIDIA that pairs the NemotronH hybrid Mamba-Transformer language model with a RADIO vision encoder","🚨 [ROCm] gpt-oss: route FA3 to aiter-flash-attn, generate ROCm fixtures (#46837)\n🚨 Speed up detr image processing (#48066)"]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, deprecation, feature, new-architecture)","llm: publish; 3 citas verificadas, 2 rechazadas"]},"sources":[{"url":"https://github.com/huggingface/transformers/releases/tag/v5.18.0","title":"huggingface/transformers Release 5.18.0","retrieved_at":"2026-10-13T05:51:18Z","sha256":"284f4ff99043cd427e11ae1afdc7a5a62858416a44cf8eb880a64614ffe1cb4e","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--huggingface--transformers--v5.18.0","related":[],"org":{"slug":"huggingface","name":"Hugging Face","kind":"empresa","url":"https://huggingface.co"},"tool":{"slug":"transformers","name":"Transformers"},"version":"v5.18.0"},{"id":"github:ggml-org/llama.cpp:b11273","kind":"software_release","title":"llama.cpp b11273","date":"2026-09-30","observed_at":"2026-10-13T05:55:14Z","subject":{"type":"repo","ref":"ggml-org/llama.cpp","url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11273","entity":"ggml","channel":"auto-build"},"status":"published","severity":"notable","action":"try","tags":["new-architecture"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":null},"highlights":[{"text":"Soporte para una nueva arquitectura de modelo","quote":"model : support classifier_pooling for rerankers (#29627)","source":0},{"text":"Soporte para una nueva arquitectura de modelo","quote":"model : support classifier_pooling for ModernBERT rerankers","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:55:14Z","rejected":[]},"triage":{"method":"rules","model":"gemma4","reasons":["reglas: publish (new-architecture)","llm lo descartó pero hay señal fuerte; se aplica la decisión de reglas"]},"sources":[{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11273","title":"ggml-org/llama.cpp b11273","retrieved_at":"2026-10-13T05:51:14Z","sha256":"37064b45072df33ae637202da3dfd9962da51a771ab1bd09ba800a63b67fc655","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ggml-org--llama.cpp--b11273","related":[],"org":{"slug":"ggml","name":"ggml.ai","kind":"proyecto","url":"https://ggml.ai"},"tool":{"slug":"llama.cpp","name":"llama.cpp"},"version":"b11273"},{"id":"rss:openai:6071f63e21f6","kind":"announcement","title":"DevDay 2026 Recap","date":"2026-09-29","observed_at":"2026-10-13T09:06:16Z","subject":{"type":"feed","ref":"openai","url":"https://openai.com/index/devday-2026-recap","entity":"openai","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":[],"affects":[],"summary":{"why":"El anuncio de una nueva generación de modelos frontera (GPT-6) y nuevas herramientas para desarrolladores obliga a evaluar el impacto en los flujos de trabajo actuales."},"highlights":[{"text":"Anuncio de GPT-6 Astra y nuevas herramientas para constructores.","quote":"including GPT-6 Astra, ChatGPT, Codex, APIs, security, and new tools for builders","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T09:06:16Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://openai.com/index/devday-2026-recap","title":"DevDay 2026 Recap","retrieved_at":"2026-10-13T09:05:01Z","sha256":"c5ec0680f0a165dadda9264ee9f05e751fc84ab32435632c069331ce8bf16f16","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--openai--6071f63e21f6","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"}},{"id":"rss:ollama-blog:bbefdaefb764","kind":"announcement","title":"Ollama now supports Jev-style decision models","date":"2026-09-29","observed_at":"2026-10-13T06:20:44Z","subject":{"type":"feed","ref":"ollama-blog","url":"https://ollama.com/blog/ollama-now-supports-jev-style-decision-models","entity":"ollama","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["feature","new-model"],"affects":["ollama"],"summary":{"why":"Ollama permite ejecutar modelos de decisión especializados (tipo Jev) que ofrecen respuestas estructuradas (sí/no, elecciones o puntuaciones) con baja latencia y sin coste adicional."},"highlights":[{"text":"Soporte para modelos de decisión basados en la API Jev de TypeSafe.","quote":"Ollama now supports decision models, based on TypeSafe's Jev API for fast, typed decisions.","source":0},{"text":"Capacidad para obtener respuestas de tipo sí/no, elecciones y puntuaciones con probabilidades.","quote":"Based on text, decision models answer yes-or-no questions, choices and scores about it, with a probability for every option.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:20:44Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 2 citas verificadas, 0 rechazadas","publish con severidad minor: queda en observación","corrección editorial: lanzamiento real nunca es minor (la severidad minor era un fallo de calibración del prompt)"]},"sources":[{"url":"https://ollama.com/blog/ollama-now-supports-jev-style-decision-models","title":"Ollama now supports Jev-style decision models","retrieved_at":"2026-10-13T05:51:54Z","sha256":"e97a3e085416cc6e0e137324fd9ac1523abe3bb29b13f0bc6b07114e0ee8b843","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--ollama-blog--bbefdaefb764","related":[],"org":{"slug":"ollama","name":"Ollama","kind":"proyecto","url":"https://ollama.com"}},{"id":"rss:openai:d6b9f5e651b1","kind":"announcement","title":"Introducing GPT-6.1 Sol","date":"2026-09-29","observed_at":"2026-10-13T06:15:27Z","subject":{"type":"feed","ref":"openai","url":"https://openai.com/index/introducing-gpt-6-1-sol","entity":"openai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":[],"affects":[],"summary":{"why":"Se introduce un nuevo modelo con capacidades de codificación y uso de computadora a un precio significativamente menor (un quinto del estándar de Astra)."},"highlights":[{"text":"Nuevo modelo GPT-6.1 Sol con capacidades de inteligencia para codificación y uso de computadora.","quote":"near-Astra intelligence for coding, computer use, and professional work","source":0},{"text":"Reducción drástica en el coste de tokens de entrada y salida.","quote":"at one-fifth of Astra’s standard API input and output token prices","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:15:27Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 2 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://openai.com/index/introducing-gpt-6-1-sol","title":"Introducing GPT-6.1 Sol","retrieved_at":"2026-10-13T05:51:53Z","sha256":"829a2eb548bb07733d12bf02b10ffdfddf336ae634415de0f4103b3ccabbd141","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--openai--d6b9f5e651b1","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"}},{"id":"hf:microsoft/AesCode-32B","kind":"model_release","title":"microsoft publica AesCode-32B","date":"2026-09-29","observed_at":"2026-10-13T06:14:49Z","subject":{"type":"model","ref":"microsoft/AesCode-32B","url":"https://huggingface.co/microsoft/AesCode-32B","entity":"microsoft","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Modelo nuevo: 33.4 B parámetros, licencia apache-2.0, 24 cuantizaciones GGUF disponibles."},"highlights":[],"facts":[{"field":"params_total","value":33357390064,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"apache-2.0","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"image-text-to-text","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-29T13:05:36.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"Qwen3VLForConditionalGeneration","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}},{"field":"gguf_quants","value":24,"evidence":{"source":2,"locator":"siblings[].size"}}],"verification":{"status":"verified","checked_at":"2026-10-13T06:14:49Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/microsoft/AesCode-32B","title":"Hugging Face · microsoft/AesCode-32B","retrieved_at":"2026-10-13T05:51:49Z","sha256":"544d221a9c0272a620e683e700c2efafd3c304f9ed819290a765567735f9dc0c","license":null},{"url":"https://huggingface.co/microsoft/AesCode-32B/blob/main/config.json","title":"microsoft/AesCode-32B · config.json","retrieved_at":"2026-10-13T05:51:49Z","sha256":"3cfa1a0d6c84f7b768ae969228a5d878f47abb91c9721d4a90a9f6fde6e3e234","license":null},{"url":"https://huggingface.co/bartowski/AesCode-32B-GGUF","title":"Hugging Face · bartowski/AesCode-32B-GGUF","retrieved_at":"2026-10-13T05:51:49Z","sha256":"1f692739f228ff81dcd90bd973eb0bf03a3cc2f5c4227db1bfb25e7739b505b2","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"microsoft/AesCode-32B","params_total":33357390064,"architecture":"Qwen3VLForConditionalGeneration","context_length":262144,"license":"apache-2.0","gated":false,"pipeline":"image-text-to-text","variants":[],"kv_bytes_per_token":262144,"quants":[{"repo":"bartowski/AesCode-32B-GGUF","quant":"IQ2_XXS","bytes":10634558272},{"repo":"bartowski/AesCode-32B-GGUF","quant":"IQ2_XS","bytes":11406572352},{"repo":"bartowski/AesCode-32B-GGUF","quant":"IQ2_S","bytes":12022810432},{"repo":"bartowski/AesCode-32B-GGUF","quant":"IQ2_M","bytes":12710938432},{"repo":"bartowski/AesCode-32B-GGUF","quant":"Q2_K","bytes":13005256512},{"repo":"bartowski/AesCode-32B-GGUF","quant":"IQ3_XXS","bytes":13596985152},{"repo":"bartowski/AesCode-32B-GGUF","quant":"IQ3_XS","bytes":14368768832},{"repo":"bartowski/AesCode-32B-GGUF","quant":"Q3_K_S","bytes":14914028352},{"repo":"bartowski/AesCode-32B-GGUF","quant":"IQ3_M","bytes":15496315712},{"repo":"bartowski/AesCode-32B-GGUF","quant":"Q3_K_M","bytes":16515728192},{"repo":"bartowski/AesCode-32B-GGUF","quant":"Q3_K_L","bytes":17519739712},{"repo":"bartowski/AesCode-32B-GGUF","quant":"IQ4_XS","bytes":18123034432},{"repo":"bartowski/AesCode-32B-GGUF","quant":"Q4_0","bytes":19049119552},{"repo":"bartowski/AesCode-32B-GGUF","quant":"IQ4_NL","bytes":19070091072},{"repo":"bartowski/AesCode-32B-GGUF","quant":"Q4_K_S","bytes":19135627072},{"repo":"bartowski/AesCode-32B-GGUF","quant":"Q4_K_M","bytes":20485668672},{"repo":"bartowski/AesCode-32B-GGUF","quant":"Q4_1","bytes":20898668352},{"repo":"bartowski/AesCode-32B-GGUF","quant":"Q4_K_L","bytes":21063025472},{"repo":"bartowski/AesCode-32B-GGUF","quant":"Q5_K_S","bytes":22813753152},{"repo":"bartowski/AesCode-32B-GGUF","quant":"Q5_K_M","bytes":23794171712},{"repo":"bartowski/AesCode-32B-GGUF","quant":"Q6_K","bytes":27980380992},{"repo":"bartowski/AesCode-32B-GGUF","quant":"Q6_K_L","bytes":28357182272},{"repo":"bartowski/AesCode-32B-GGUF","quant":"Q8_0","bytes":34817721152},{"repo":"bartowski/AesCode-32B-GGUF","quant":"BF16","bytes":65531577056}]},"slug":"hf--microsoft--aescode-32b","related":[],"org":{"slug":"microsoft","name":"Microsoft","kind":"empresa","url":"https://www.microsoft.com"}},{"id":"hf:microsoft/AesCode-8B","kind":"model_release","title":"microsoft publica AesCode-8B","date":"2026-09-29","observed_at":"2026-10-13T06:14:49Z","subject":{"type":"model","ref":"microsoft/AesCode-8B","url":"https://huggingface.co/microsoft/AesCode-8B","entity":"microsoft","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Modelo nuevo: 8.8 B parámetros, licencia apache-2.0, 22 cuantizaciones GGUF disponibles."},"highlights":[],"facts":[{"field":"params_total","value":8767123696,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"apache-2.0","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"image-text-to-text","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-29T13:04:44.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"Qwen3VLForConditionalGeneration","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}},{"field":"gguf_quants","value":22,"evidence":{"source":2,"locator":"siblings[].size"}}],"verification":{"status":"verified","checked_at":"2026-10-13T06:14:49Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/microsoft/AesCode-8B","title":"Hugging Face · microsoft/AesCode-8B","retrieved_at":"2026-10-13T05:51:49Z","sha256":"fd5cf5db525f732a34a364140b7ee7e2fd8ad9892db284d19dc40b3bcc0600ea","license":null},{"url":"https://huggingface.co/microsoft/AesCode-8B/blob/main/config.json","title":"microsoft/AesCode-8B · config.json","retrieved_at":"2026-10-13T05:51:50Z","sha256":"aa2412f4138e442a31824766489ff39cd1df16ce4bd5ca8221518f60558e5e7d","license":null},{"url":"https://huggingface.co/bartowski/AesCode-8B-GGUF","title":"Hugging Face · bartowski/AesCode-8B-GGUF","retrieved_at":"2026-10-13T05:51:50Z","sha256":"88113e1a5d95c498bd4e1a9bcc9ceb2e22d090c1eb895da4029cb44f503dfd4d","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"microsoft/AesCode-8B","params_total":8767123696,"architecture":"Qwen3VLForConditionalGeneration","context_length":262144,"license":"apache-2.0","gated":false,"pipeline":"image-text-to-text","variants":[],"kv_bytes_per_token":147456,"quants":[{"repo":"bartowski/AesCode-8B-GGUF","quant":"IQ2_M","bytes":3139898336},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q2_K","bytes":3299576800},{"repo":"bartowski/AesCode-8B-GGUF","quant":"IQ3_XXS","bytes":3684383712},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q3_K_S","bytes":3870899168},{"repo":"bartowski/AesCode-8B-GGUF","quant":"IQ3_XS","bytes":3886431200},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q3_K_M","bytes":4039424992},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q3_K_L","bytes":4230691808},{"repo":"bartowski/AesCode-8B-GGUF","quant":"IQ3_M","bytes":4475763680},{"repo":"bartowski/AesCode-8B-GGUF","quant":"IQ4_XS","bytes":4688030688},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q4_0","bytes":4865190880},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q4_K_S","bytes":4928039904},{"repo":"bartowski/AesCode-8B-GGUF","quant":"IQ4_NL","bytes":5178452960},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q4_K_M","bytes":5211515872},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q4_1","bytes":5306739680},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q4_K_L","bytes":5586119648},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q5_K_S","bytes":5891845088},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q5_K_M","bytes":6242790368},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q6_K_S","bytes":6857956320},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q6_K","bytes":7140350944},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q6_K_L","bytes":7457283040},{"repo":"bartowski/AesCode-8B-GGUF","quant":"Q8_0","bytes":8709520352},{"repo":"bartowski/AesCode-8B-GGUF","quant":"BF16","bytes":16388045504}]},"slug":"hf--microsoft--aescode-8b","related":[],"org":{"slug":"microsoft","name":"Microsoft","kind":"empresa","url":"https://www.microsoft.com"}},{"id":"github:NVIDIA/TensorRT-LLM:v1.3.0rc29","kind":"software_release","title":"TensorRT-LLM v1.3.0rc29","date":"2026-09-29","observed_at":"2026-10-13T06:12:59Z","subject":{"type":"repo","ref":"NVIDIA/TensorRT-LLM","url":"https://github.com/NVIDIA/TensorRT-LLM/releases/tag/v1.3.0rc29","entity":"nvidia","channel":"auto-build"},"status":"published","severity":"notable","action":"watch","tags":["breaking","performance"],"affects":["tensorrt-llm"],"summary":{"why":"Esta versión incluye cambios disruptivos (breaking changes) en la API y soporte optimizado para nuevos modelos y arquitecturas (Qwen3.5, DeepSeek V4, Gemma4) mediante cuantización FP8/FP4."},"highlights":[{"text":"Cambios disruptivos en la API y eliminación de opciones de build.","quote":"Remove the AutoDeploy integration and its public entry points (BREAKING)","source":0},{"text":"Soporte para checkpoints de Qwen3.5 con escalas FP8 globales.","quote":"Support Qwen3.5 checkpoints with global FP8 scales across model loading paths #19519","source":0},{"text":"Optimización de embeddings rotatorios para la visión de Gemma4.","quote":"Optimize Gemma4 vision rotary embeddings without a GEMM operation #19571","source":0},{"text":"Soporte para NVFP4 MLA residual switch para DeepSeek V4.","quote":"Add an NVFP4 MLA residual switch for DeepSeek V4 #19305","source":0},{"text":"Actualización de dependencias críticas (PyTorch y Triton).","quote":"Upgrade public PyTorch to 2.14, Triton to 3.8, and C++ to 20 #19477, #19249","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:12:59Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, deprecation)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/NVIDIA/TensorRT-LLM/releases/tag/v1.3.0rc29","title":"NVIDIA/TensorRT-LLM v1.3.0rc29","retrieved_at":"2026-10-13T05:51:39Z","sha256":"29cffbd1f911487946b3d546f07cdd0af418cda8af4f21e24227b274402f76fd","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--nvidia--tensorrt-llm--v1.3.0rc29","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"},"tool":{"slug":"tensorrt-llm","name":"TensorRT-LLM"},"version":"v1.3.0rc29"},{"id":"github:ml-explore/mlx:v0.32.3","kind":"software_release","title":"mlx v0.32.3","date":"2026-09-29","observed_at":"2026-10-13T06:12:31Z","subject":{"type":"repo","ref":"ml-explore/mlx","url":"https://github.com/ml-explore/mlx/releases/tag/v0.32.3","entity":"mlx","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["hardware","performance"],"affects":["mlx"],"summary":{"why":"Esta actualización incluye optimizaciones críticas para Metal (Apple Silicon) y CUDA, además de mejoras de rendimiento en kernels de atención y soporte para hardware específico como M5 Ultra."},"highlights":[{"text":"Optimizaciones de rendimiento para M5 Ultra en multiplicaciones de matrices no cuantizadas.","quote":"Add M5 ultra tunings for non-quantized matmuls by @jagrit06 in https://github.com/ml-explore/mlx/pull/4447","source":0},{"text":"Mejoras en el uso de memoria para SDPA (Scaled Dot Product Attention) en Metal.","quote":"Improve metal memory usage for SDPA D256 by @dhiltgen in https://github.com/ml-explore/mlx/pull/4505","source":0},{"text":"Nuevos kernels de Metal fusionados para cross_entropy rápida.","quote":"Add fused Metal kernels for fast.cross_entropy by @zsun6 in https://github.com/ml-explore/mlx/pull/4520","source":0}],"facts":[],"verification":{"status":"partial","checked_at":"2026-10-13T06:12:31Z","rejected":["Add D512 support to Metal vector attention by @wyanzhao in https://github.com/ml-explore/mlx/pull/4559","Win: Use DXGI for accurate WDDM VRAM memory budget by @dhiltgen in https://github.com/ml-explore/mlx/pull/4557"]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: drop (sin señales)","llm: publish; 3 citas verificadas, 2 rechazadas"]},"sources":[{"url":"https://github.com/ml-explore/mlx/releases/tag/v0.32.3","title":"ml-explore/mlx v0.32.3","retrieved_at":"2026-10-13T05:51:37Z","sha256":"8ab992dde89d4481c358113eef7ee0b701c64aa7143f16fa027c4853ba5868bd","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ml-explore--mlx--v0.32.3","related":[],"org":{"slug":"mlx","name":"MLX (Apple)","kind":"proyecto","url":"https://github.com/ml-explore/mlx"},"tool":{"slug":"mlx","name":"MLX"},"version":"v0.32.3"},{"id":"github:openai/codex:rust-v0.158.0","kind":"software_release","title":"codex 0.158.0","date":"2026-09-28","observed_at":"2026-10-13T06:06:39Z","subject":{"type":"repo","ref":"openai/codex","url":"https://github.com/openai/codex/releases/tag/rust-v0.158.0","entity":"openai","channel":"stable"},"status":"published","severity":"notable","action":"try","tags":["feature","security"],"affects":["codex"],"summary":{"why":"Esta actualización introduce mejoras críticas de seguridad mediante la autenticación por tokens en conexiones WebSocket y soporte para secretos OAuth en servidores MCP, además de nuevas capacidades de control de permisos en la terminal."},"highlights":[{"text":"Soporte para secretos de cliente OAuth en servidores MCP.","quote":"Connect to MCP servers that require pre-registered OAuth client secrets, including through `codex mcp add --oauth-client-secret`.","source":0},{"text":"Conexiones WebSocket seguras con tokens de portador (bearer tokens).","quote":"Secure direct exec-server WebSocket connections with bearer tokens, including connections configured through app-server.","source":0},{"text":"Aprobación de entrada de terminal habilitada por defecto para permisos elevados.","quote":"Terminal input approval is enabled by default for commands running with elevated permissions","source":0},{"text":"Mejoras en la generación de imágenes con soporte para fondos transparentes.","quote":"Image generation and editing can explicitly request transparent backgrounds","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:06:39Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: drop (sin señales)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/openai/codex/releases/tag/rust-v0.158.0","title":"openai/codex 0.158.0","retrieved_at":"2026-10-13T05:51:28Z","sha256":"276eb34cc86bc679253128686766d05fbe4ebb1cd3bc6fcd8b323fc43a698518","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--openai--codex--rust-v0.158.0","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"},"tool":{"slug":"codex","name":"Codex CLI"},"version":"rust-v0.158.0"},{"id":"github:anthropics/claude-code:v2.1.284","kind":"software_release","title":"claude-code v2.1.284","date":"2026-09-28","observed_at":"2026-10-13T05:57:17Z","subject":{"type":"repo","ref":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.284","entity":"anthropic","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["feature","new-model"],"affects":["claude-code"],"summary":{"why":"Actualización importante que introduce Claude 3.5 Sonnet como modelo por defecto con soporte para 1M de contexto y mejoras significativas en la gestión de errores, MCP y modo auto."},"highlights":[{"text":"Soporte para Claude 3.5 Sonnet con 1M de contexto","quote":"Added Claude Sonnet 5.5 (`claude-sonnet-5-5`), now the default Sonnet model on the Anthropic API — 1M context, $2/$10 per Mtok with $0.20/Mtok cache reads","source":0},{"text":"Mejoras en la reconexión de servidores MCP","quote":"Added `/mcp reconnect all` in the interactive terminal to retry every MCP server that failed to connect or needs authentication at once","source":0},{"text":"Corrección de errores de compresión de prompts","quote":"Fixed \"Prompt is too long\" errors that persisted after compacting: when the compacted request is still too long, Claude Code now compacts once more, keeping less of the recent conversation","source":0},{"text":"Cambio en el modo de permisos por defecto","quote":"Changed interactive terminal and VS Code sessions to start in auto mode when no permission mode is configured","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:57:17Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, feature)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.284","title":"anthropics/claude-code v2.1.284","retrieved_at":"2026-10-13T05:51:19Z","sha256":"99d419ca7ae28983492378ca9d297989a92ec935cf25c1d620cc4e323239875e","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--anthropics--claude-code--v2.1.284","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"},"tool":{"slug":"claude-code","name":"Claude Code"},"version":"v2.1.284"},{"id":"github:ollama/ollama:v0.35.0","kind":"software_release","title":"ollama v0.35.0","date":"2026-09-28","observed_at":"2026-10-13T05:55:41Z","subject":{"type":"repo","ref":"ollama/ollama","url":"https://github.com/ollama/ollama/releases/tag/v0.35.0","entity":"ollama","channel":"stable"},"status":"published","severity":"notable","action":"try","tags":["api","new-architecture"],"affects":["ollama"],"summary":{"why":"Ollama introduce un nuevo endpoint para modelos de decisión que permite obtener probabilidades y puntuaciones en lugar de texto plano, ideal para tareas de clasificación y enrutamiento."},"highlights":[{"text":"Soporte para modelos de decisión mediante el nuevo endpoint /v1/systemone.","quote":"Ollama now supports decision models through `/v1/systemone`","source":0},{"text":"Los modelos de decisión devuelven elecciones, probabilidades y puntuaciones.","quote":"Decision models return choices, probabilities, and scores instead of text.","source":0},{"text":"Soporte para tres tipos de preguntas: choice, noul y score.","quote":"The API supports three question types: `choice`: Select an option and return probabilities for each. `noul`: Return the probability that a condition is true. `score`: Return a score across an ordered set of criteria","source":0},{"text":"Corrección de errores en descargas de modelos MLX que se quedaban colgadas.","quote":"Fixed stalled MLX model downloads hanging indefinitely.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:55:41Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (deprecation)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/ollama/ollama/releases/tag/v0.35.0","title":"ollama/ollama v0.35.0","retrieved_at":"2026-10-13T05:51:16Z","sha256":"8d6ea6f9579c2af3c4eed14846940e070a786ba2c3523b8fca01ec3b117b4a83","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ollama--ollama--v0.35.0","related":[],"org":{"slug":"ollama","name":"Ollama","kind":"proyecto","url":"https://ollama.com"},"tool":{"slug":"ollama","name":"Ollama"},"version":"v0.35.0"},{"id":"github:LostRuins/koboldcpp:v1.122.1","kind":"software_release","title":"koboldcpp-1.122.1","date":"2026-09-26","observed_at":"2026-10-13T06:14:21Z","subject":{"type":"repo","ref":"LostRuins/koboldcpp","url":"https://github.com/LostRuins/koboldcpp/releases/tag/v1.122.1","entity":"koboldcpp","channel":"stable"},"status":"published","severity":"notable","action":"try","tags":["new-architecture","new-model"],"affects":["koboldcpp"],"summary":{"why":"Introduce un agente integrado capaz de ejecutar herramientas (agentic harness) que compite con herramientas como Claude Code o OpenCode, permitiendo la integración de MCP tools."},"highlights":[{"text":"Nuevo agente integrado con 9 herramientas predefinidas y soporte para MCP.","quote":"NEW: Added an integrated KoboldCpp Agent - KoboldCpp now ships with a bundled Agentic harness. This works like an extremely lightweight replacement for tools like Opencode, Codex or Claude Code.","source":0},{"text":"Soporte para herramientas MCP mediante archivos mcp.json.","quote":"Add more tools by loading a mcp.json file, MCP tools will be shared to the agent.","source":0},{"text":"Cambio importante (breaking change) en el script de construcción para mantenedores de paquetes.","quote":"This release includes a breaking change to koboldcpp.sh. To shorten compiles and make it easier to build from source, koboldcpp.sh now attempts to optimize for the target platform it is run on.","source":0},{"text":"Nueva funcionalidad para controlar el umbral de autoswap.","quote":"Added `--autoswapthreshold`, now you can set the size limit before autoswap triggers to prevent swapping when small models are being utilized.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:14:21Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, feature)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/LostRuins/koboldcpp/releases/tag/v1.122.1","title":"LostRuins/koboldcpp koboldcpp-1.122.1","retrieved_at":"2026-10-13T05:51:44Z","sha256":"aa70ea382f30880dfc93f1a14e687b2d61671b52c8407c5f2ecbe05335a935bf","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--lostruins--koboldcpp--v1.122.1","related":[],"org":{"slug":"koboldcpp","name":"KoboldCpp","kind":"proyecto","url":"https://koboldai.org"},"tool":{"slug":"koboldcpp","name":"KoboldCpp"},"version":"v1.122.1"},{"id":"github:openai/codex:rust-v0.157.0","kind":"software_release","title":"codex 0.157.0","date":"2026-09-25","observed_at":"2026-10-13T06:07:31Z","subject":{"type":"repo","ref":"openai/codex","url":"https://github.com/openai/codex/releases/tag/rust-v0.157.0","entity":"openai","channel":"stable"},"status":"published","severity":"notable","action":"try","tags":["feature","new-model"],"affects":["codex"],"summary":{"why":"La actualización introduce soporte para nuevos modelos (GPT-6 Sol y Luna) y compatibilidad con Amazon Bedrock, además de mejoras significativas en la gestión de sesiones y red."},"highlights":[{"text":"Inclusión de nuevos modelos GPT-6 Sol y Luna con soporte para Amazon Bedrock.","quote":"Added GPT-6 Sol and Luna, including Amazon Bedrock support and migration prompts for older models.","source":0},{"text":"Mejoras en la interfaz de terminal (TUI) con renderizado de ecuaciones y listas Markdown.","quote":"Improved terminal rendering with Unicode bullets, checkboxes, aligned equations, and optimization notation.","source":0},{"text":"Nuevas capacidades de importación en sesiones remotas y locales.","quote":"Made `/import` available in remote sessions and local background-server sessions.","source":0},{"text":"Aumento del timeout para la subida de archivos de OpenAI a 5 minutos.","quote":"increased the upload timeout to five minutes.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:07:31Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (deprecation, feature)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/openai/codex/releases/tag/rust-v0.157.0","title":"openai/codex 0.157.0","retrieved_at":"2026-10-13T05:51:28Z","sha256":"4753c15d7453de31b2a76aa2150f8144de122105d8fd45dac698c7f4f5e35f08","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--openai--codex--rust-v0.157.0","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"},"tool":{"slug":"codex","name":"Codex CLI"},"version":"rust-v0.157.0"},{"id":"github:ollama/ollama:v0.40.0","kind":"software_release","title":"ollama v0.40.0","date":"2026-09-25","observed_at":"2026-10-13T05:55:32Z","subject":{"type":"repo","ref":"ollama/ollama","url":"https://github.com/ollama/ollama/releases/tag/v0.40.0","entity":"ollama","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["hardware","performance"],"affects":["ollama"],"summary":{"why":"Optimiza drásticamente el rendimiento en dispositivos Apple Silicon al cambiar el backend por defecto a MLX para arquitecturas compatibles."},"highlights":[{"text":"Los modelos en Apple Silicon ahora usan el runtime de MLX por defecto.","quote":"In this release, on Apple Silicon devices, model architectures supported by the MLX runtime will automatically run on MLX.","source":0}],"facts":[],"verification":{"status":"partial","checked_at":"2026-10-13T05:55:32Z","rejected":["Additional models include gemma4, qwen3.6 and qwen3.5","Decision models are now available on MLX as well: Nimble tev1 clef clef-flash"]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: drop (sin señales)","llm: publish; 1 citas verificadas, 2 rechazadas"]},"sources":[{"url":"https://github.com/ollama/ollama/releases/tag/v0.40.0","title":"ollama/ollama v0.40.0","retrieved_at":"2026-10-13T05:51:16Z","sha256":"8309c257ab0eefb6edd9d5fd829a22519c6839345571cb71de314946edfdd7ea","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--ollama--ollama--v0.40.0","related":[],"org":{"slug":"ollama","name":"Ollama","kind":"proyecto","url":"https://ollama.com"},"tool":{"slug":"ollama","name":"Ollama"},"version":"v0.40.0"},{"id":"rss:nvidia-dev:9c631b921733","kind":"announcement","title":"Introducing NV-Reason-CT Open 3D CT VLM for Radiologist Chain-of-Thought Reasoning","date":"2026-09-23","observed_at":"2026-10-13T06:19:20Z","subject":{"type":"feed","ref":"nvidia-dev","url":"https://developer.nvidia.com/blog/introducing-nv-reason-ct-open-3d-ct-vlm-for-radiologist-chain-of-thought-reasoning/","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":["pytorch"],"summary":{"why":"NVIDIA presenta un nuevo modelo de visión-lenguaje (VLM) especializado en razonamiento 3D para tomografías computarizadas (CT), orientado a tareas médicas complejas."},"highlights":[{"text":"Lanzamiento de NV-Reason-CT, un modelo de visión-lenguaje para razonamiento en cadena de pensamiento (CoT) sobre imágenes 3D de CT.","quote":"Introducing NV-Reason-CT Open 3D CT VLM for Radiologist Chain-of-Thought Reasoning","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:19:20Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas","publish con severidad minor: queda en observación","corrección editorial: lanzamiento real nunca es minor (la severidad minor era un fallo de calibración del prompt)"]},"sources":[{"url":"https://developer.nvidia.com/blog/introducing-nv-reason-ct-open-3d-ct-vlm-for-radiologist-chain-of-thought-reasoning/","title":"Introducing NV-Reason-CT Open 3D CT VLM for Radiologist Chain-of-Thought Reasoning","retrieved_at":"2026-10-13T05:51:54Z","sha256":"0f002a1043beb22dd082cd86f7ea29cd8450e30cd138114fb7cd0fecc28619ab","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--nvidia-dev--9c631b921733","related":["hf:nvidia/NV-Reason-CT"],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}},{"id":"github:NVIDIA/TensorRT-LLM:v1.3.0rc28","kind":"software_release","title":"TensorRT-LLM v1.3.0rc28","date":"2026-09-23","observed_at":"2026-10-13T06:13:04Z","subject":{"type":"repo","ref":"NVIDIA/TensorRT-LLM","url":"https://github.com/NVIDIA/TensorRT-LLM/releases/tag/v1.3.0rc28","entity":"nvidia","channel":"auto-build"},"status":"published","severity":"notable","action":"watch","tags":["breaking","performance"],"affects":["tensorrt-llm"],"summary":{"why":"Esta versión incluye cambios de ruptura (breaking changes) en la API y activa por defecto el nuevo gestor de KV-cache para modelos Llama, lo que requiere revisión antes de actualizar en entornos de producción."},"highlights":[{"text":"Cambios de ruptura en la API que eliminan rutas de evaluación y benchmark de TensorRT.","quote":"Remove deprecated TensorRT serve, evaluation, benchmark, and stress paths (BREAKING)","source":0},{"text":"Activación por defecto del KV-cache manager V2 para la familia Llama.","quote":"Enable KV-cache manager V2 by default for Llama and Llama4","source":0},{"text":"Soporte extendido para modelos Qwen3.8-Flash-Next y MiniMax-M3.","quote":"Extend Qwen3.8-Flash-Next with disaggregation, MegaMoE, and GVR index reuse","source":0},{"text":"Introducción de soporte para DFlash 2.","quote":"Add DFlash 2 support with model-specific MoE backend overrides","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:13:04Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, deprecation)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/NVIDIA/TensorRT-LLM/releases/tag/v1.3.0rc28","title":"NVIDIA/TensorRT-LLM v1.3.0rc28","retrieved_at":"2026-10-13T05:51:39Z","sha256":"ff1fd3bac1034888cc4a63faff3d86563fd3aa4a30c92ff0582fc80d152d5e12","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--nvidia--tensorrt-llm--v1.3.0rc28","related":["hf:Qwen/Qwen3.8-Flash-Next"],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"},"tool":{"slug":"tensorrt-llm","name":"TensorRT-LLM"},"version":"v1.3.0rc28"},{"id":"github:anthropics/claude-code:v2.1.281","kind":"software_release","title":"claude-code v2.1.281","date":"2026-09-23","observed_at":"2026-10-13T05:57:36Z","subject":{"type":"repo","ref":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.281","entity":"anthropic","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["fix","security"],"affects":["claude-code"],"summary":{"why":"Esta actualización corrige fallos críticos de seguridad en el manejo de permisos (como la ejecución de `rm` no solicitada) y errores graves de estabilidad en la gestión de sesiones y herramientas MCP."},"highlights":[{"text":"Corrección de seguridad en comandos recursivos `rm` que se ejecutaban sin permiso en modo auto.","quote":"Fixed a recursive `rm` whose target is only command-substitution output, such as `rm -rf \"$(pwd)\"`, running unprompted in auto and `--dangerously-skip-permissions` mode","source":0},{"text":"Mejoras en el soporte de Bedrock con roles IAM y guardrails de Amazon.","quote":"Added `assume_role` on Claude apps gateway Bedrock upstreams","source":0},{"text":"Corrección de errores de estabilidad al reanudar sesiones y manejar el caché de prompts.","quote":"Fixed resuming a session that ended during a tool call: Claude now sees the call and is told its outcome is unknown","source":0},{"text":"Mejoras en la validación de plugins MCP para detectar entradas inseguras.","quote":"Added MCP server checks to `claude plugin validate`: it reports .mcp.json entries that would be silently dropped at load, undeclared ${user_config.*} references, and insecure URLs","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:57:36Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, feature)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.281","title":"anthropics/claude-code v2.1.281","retrieved_at":"2026-10-13T05:51:19Z","sha256":"e70dc58d1dbd8ae15c4083490655fcc4a93ed3f3cf738d1151aa0d97cb3e536e","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--anthropics--claude-code--v2.1.281","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"},"tool":{"slug":"claude-code","name":"Claude Code"},"version":"v2.1.281"},{"id":"rss:hf-blog:689d78c6e007","kind":"announcement","title":"Transformers now runs llama.cpp quants","date":"2026-09-22","observed_at":"2026-10-13T06:18:17Z","subject":{"type":"feed","ref":"hf-blog","url":"https://huggingface.co/blog/transformers-llama-cpp-quants","entity":"huggingface","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model","performance"],"affects":["llama.cpp","transformers"],"summary":{"why":"La integración de cuantizaciones de llama.cpp en la librería Transformers permite cargar modelos optimizados directamente, facilitando el uso de pesos comprimidos sin conversiones manuales complejas."},"highlights":[{"text":"Soporte para cuantizaciones de llama.cpp en la librería Transformers","quote":"Transformers now runs llama.cpp quants","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:18:17Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://huggingface.co/blog/transformers-llama-cpp-quants","title":"Transformers now runs llama.cpp quants","retrieved_at":"2026-10-13T05:51:54Z","sha256":"6be58508088242cdcc1941fa933a9f53524c8ebba29a3aafe1140ea6750cfa71","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--hf-blog--689d78c6e007","related":[],"org":{"slug":"huggingface","name":"Hugging Face","kind":"empresa","url":"https://huggingface.co"}},{"id":"rss:openai:73461b3e97af","kind":"announcement","title":"Introducing GPT-6 Sol and Luna","date":"2026-09-22","observed_at":"2026-10-13T06:16:08Z","subject":{"type":"feed","ref":"openai","url":"https://openai.com/index/introducing-gpt-6-sol-and-luna","entity":"openai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":[],"summary":{"why":"OpenAI lanza dos nuevos modelos de la familia GPT-6 con diferentes perfiles de capacidad y coste, lo que implica nuevas opciones de inteligencia de frontera para integración."},"highlights":[{"text":"Presentación de GPT-6 Sol y Luna, dos modelos con diferentes equilibrios entre capacidad y coste.","quote":"Meet GPT-6 Sol and Luna, two models that bring frontier intelligence to everyday work with different balances of capability and cost.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:16:08Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://openai.com/index/introducing-gpt-6-sol-and-luna","title":"Introducing GPT-6 Sol and Luna","retrieved_at":"2026-10-13T05:51:53Z","sha256":"881bdf7375f4fad6f2ddac3add4794134536f3ec7d9ee56e1dcae71f18e31338","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--openai--73461b3e97af","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"}},{"id":"rss:openai:f9027d80e820","kind":"announcement","title":"Better prompt caching for GPT-6","date":"2026-09-22","observed_at":"2026-10-13T06:16:05Z","subject":{"type":"feed","ref":"openai","url":"https://openai.com/index/better-prompt-caching-for-gpt-6","entity":"openai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":[],"affects":[],"summary":{"why":"La mejora en la tasa de aciertos de caché y los nuevos controles de latencia/costes impactan directamente en la eficiencia de aplicaciones que utilizan modelos de OpenAI."},"highlights":[{"text":"Mejoras en la tasa de aciertos de caché de prompts","quote":"higher cache hit rates","source":0},{"text":"Nuevos diagnósticos y puntos de interrupción explícitos","quote":"new diagnostics, explicit breakpoints","source":0},{"text":"Controles para reducir latencia y costes","quote":"controls that reduce latency and costs","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:16:05Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 3 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://openai.com/index/better-prompt-caching-for-gpt-6","title":"Better prompt caching for GPT-6","retrieved_at":"2026-10-13T05:51:53Z","sha256":"08d420ebe69764945ef16e63d1b304d3be7cb2422d8189aafec6faf23a125aa5","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--openai--f9027d80e820","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"}},{"id":"github:anthropics/claude-code:v2.1.280","kind":"software_release","title":"claude-code v2.1.280","date":"2026-09-22","observed_at":"2026-10-13T05:57:41Z","subject":{"type":"repo","ref":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.280","entity":"anthropic","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":[],"affects":["claude-code"],"summary":{"why":"Actualización importante que introduce un nuevo modelo por defecto (Claude Opus 5.5) con mayor ventana de contexto y cambios en la estructura de costes y configuración de esfuerzo."},"highlights":[{"text":"Nuevo modelo Claude Opus 5.5 por defecto con 1M de contexto.","quote":"Added Claude Opus 5.5 (`claude-opus-5-5`), now the default Opus model — 1M context","source":0},{"text":"Cambio en el modelo por defecto para planes Pro y Team Standard.","quote":"Changed the default model on Pro and Team Standard plans from Sonnet to Opus","source":0},{"text":"Nueva variable para ajustar el límite de descripción de herramientas MCP.","quote":"Added `CLAUDE_CODE_MAX_MCP_DESCRIPTION_LENGTH` to change the 2,048-character cap on MCP tool descriptions","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:57:41Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (feature)","llm: publish; 3 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.280","title":"anthropics/claude-code v2.1.280","retrieved_at":"2026-10-13T05:51:19Z","sha256":"33a97d673bed299fa0b8d734cbef2258c0c6b5c3135c2e26d3115a514884e911","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--anthropics--claude-code--v2.1.280","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"},"tool":{"slug":"claude-code","name":"Claude Code"},"version":"v2.1.280"},{"id":"github:vllm-project/vllm:v0.30.0","kind":"software_release","title":"vllm v0.30.0","date":"2026-09-22","observed_at":"2026-10-13T05:56:23Z","subject":{"type":"repo","ref":"vllm-project/vllm","url":"https://github.com/vllm-project/vllm/releases/tag/v0.30.0","entity":"vllm","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["breaking","new-model","performance"],"affects":["vllm"],"summary":{"why":"Es una actualización mayor con soporte para modelos críticos (DeepSeek-V4, Qwen3.8, Gemma 4), mejoras masivas de rendimiento en kernels y cambios que rompen la compatibilidad en el despliegue de endpoints."},"highlights":[{"text":"Mejora en el tiempo de inicio mediante un demonio de caché de pesos por GPU.","quote":"a persistent per-GPU weight-cache daemon holds post-quantized, TP-sharded weights in GPU memory so restarting engines map them over CUDA IPC with `--load-format ipc_cache` instead of reloading from disk","source":0},{"text":"Nuevas capacidades de watermarking para la generación de texto.","quote":"Gumbel-max watermarked generation and detection with a keyed PRF, per-request opt-out and an example detection endpoint","source":0},{"text":"Cambios que rompen la compatibilidad (Breaking changes) en la configuración de endpoints y eliminación de variables de entorno.","quote":"scale-out endpoints are opt-in on plain `vllm serve` via `--enable-scale-out`, replacing `VLLM_ENABLE_SCALE_OUT_ENDPOINTS`","source":0},{"text":"Optimización de rendimiento para modelos multimodales y soporte de audio con torchcodec.","quote":"torchcodec audio decoding and a selectable `audio_backend` in `--media-io-kwargs`","source":0}],"facts":[],"verification":{"status":"partial","checked_at":"2026-10-13T05:56:23Z","rejected":["New models: DeepSeek-V4.1-Flash (#56214, #56228, #56208) ... GLM-5.3-Flash (#53906) ... Kimi K3 performance"]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, deprecation, performance)","llm: publish; 4 citas verificadas, 1 rechazadas"]},"sources":[{"url":"https://github.com/vllm-project/vllm/releases/tag/v0.30.0","title":"vllm-project/vllm v0.30.0","retrieved_at":"2026-10-13T05:51:17Z","sha256":"05a90d62703d42309b6177ac5bf1a2799469609e440ae257516f10c77769ed0b","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--vllm-project--vllm--v0.30.0","related":[],"org":{"slug":"vllm","name":"vLLM","kind":"proyecto","url":"https://vllm.ai"},"tool":{"slug":"vllm","name":"vLLM"},"version":"v0.30.0"},{"id":"hf:microsoft/rho-roboeval","kind":"model_release","title":"microsoft publica rho-roboeval","date":"2026-09-21","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"microsoft/rho-roboeval","url":"https://huggingface.co/microsoft/rho-roboeval","entity":"microsoft","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 5.2 B parámetros, licencia mit, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":5238067680,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"mit","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"created_at","value":"2026-09-21T19:13:53.000Z","evidence":{"source":0,"locator":"createdAt"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/microsoft/rho-roboeval","title":"Hugging Face · microsoft/rho-roboeval","retrieved_at":"2026-10-13T08:54:55Z","sha256":"4025c047872694345ac3c81060a7091583a90d055d079ee89dac1489d23c0adb","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"microsoft/rho-roboeval","params_total":5238067680,"architecture":null,"context_length":null,"license":"mit","gated":false,"pipeline":null,"variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--microsoft--rho-roboeval","related":[],"org":{"slug":"microsoft","name":"Microsoft","kind":"empresa","url":"https://www.microsoft.com"}},{"id":"rss:xai-news:39b901174b54","kind":"announcement","title":"Introducing Grok 4.7 | SpaceXSI","date":"2026-09-21","observed_at":"2026-10-13T06:23:22Z","subject":{"type":"feed","ref":"xai-news","url":"https://x.ai/news/grok-4-7","entity":"xai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":[],"summary":{"why":"Es un nuevo modelo orientado a tareas de programación y conocimiento con mejoras significativas de velocidad y coste."},"highlights":[{"text":"Modelo optimizado para codificación y trabajo de conocimiento.","quote":"most powerful model for coding and knowledge work","source":0},{"text":"Doble de velocidad que modelos comparables.","quote":"Twice as fast","source":0},{"text":"Mitad de precio que modelos comparables.","quote":"at half the price of comparable models","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:23:22Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 3 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://x.ai/news/grok-4-7","title":"Introducing Grok 4.7 | SpaceXSI","retrieved_at":"2026-10-13T05:52:15Z","sha256":"6d457e8b030ec37c74adf5e610f53280edcf81f665522a16fb1a41b0d3a0dcb8","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--xai-news--39b901174b54","related":[],"org":{"slug":"xai","name":"xAI","kind":"laboratorio","url":"https://x.ai"}},{"id":"rss:nvidia-dev:9ee29b369590","kind":"announcement","title":"Simplifying Model Serving Across Multiple GPUs with NVIDIA TensorRT Multi-Device Integration in NVIDIA Dynamo-Triton","date":"2026-09-21","observed_at":"2026-10-13T06:19:34Z","subject":{"type":"feed","ref":"nvidia-dev","url":"https://developer.nvidia.com/blog/simplifying-model-serving-across-multiple-gpus-with-nvidia-tensorrt-multi-device-integration-in-nvidia-dynamo-triton/","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":[],"affects":["tensorrt-llm"],"summary":{"why":"Permite realizar inferencia distribuida en múltiples GPUs para modelos que superan la capacidad de memoria de una sola tarjeta, optimizando el serving mediante la integración de TensorRT en Dynamo-Triton."},"highlights":[{"text":"Nueva capacidad de inferencia multi-dispositivo con NVIDIA TensorRT para superar las limitaciones de memoria y cómputo de una sola GPU.","quote":"NVIDIA TensorRT multi-device inference is a new capability","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:19:34Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://developer.nvidia.com/blog/simplifying-model-serving-across-multiple-gpus-with-nvidia-tensorrt-multi-device-integration-in-nvidia-dynamo-triton/","title":"Simplifying Model Serving Across Multiple GPUs with NVIDIA TensorRT Multi-Device Integration in NVIDIA Dynamo-Triton","retrieved_at":"2026-10-13T05:51:54Z","sha256":"e37cf3507d99f1d44d17c3d1f8b8e19b151b0425c85f51f1f1582b983a405963","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--nvidia-dev--9ee29b369590","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}},{"id":"github:open-webui/open-webui:v0.11.4","kind":"software_release","title":"open-webui v0.11.4","date":"2026-09-21","observed_at":"2026-10-13T06:14:11Z","subject":{"type":"repo","ref":"open-webui/open-webui","url":"https://github.com/open-webui/open-webui/releases/tag/v0.11.4","entity":"open-webui","channel":"auto-build"},"status":"published","severity":"notable","action":"try","tags":["feature","new-model"],"affects":["open-webui"],"summary":{"why":"Esta actualización reduce drásticamente el tamaño de las imágenes Docker y añade capacidades avanzadas de terminal y gestión de 'skills' para agentes, mejorando la eficiencia de despliegue y la interacción con el sistema."},"highlights":[{"text":"Optimización de la imagen estándar eliminando duplicados.","quote":"The image no longer carries a second copy of Python, two sets of fonts nothing ever loaded, packages nothing imports, or the tool that installed them, taking about 170 MB off a standard build.","source":0},{"text":"Integración de habilidades (skills) directamente desde la terminal.","quote":"Skills a connected terminal server offers now sit beside workspace skills everywhere skills are picked","source":0},{"text":"Descubrimiento automático de habilidades para el modelo.","quote":"Every skill you can reach is now listed to the model by name and description","source":0},{"text":"Soporte para archivos de instrucciones AGENTS.md en terminales.","quote":"A model working with a terminal is now handed the AGENTS.md sitting in that terminal's home directory","source":0}],"facts":[],"verification":{"status":"partial","checked_at":"2026-10-13T06:14:11Z","rejected":["A slim build now comes down to around 175 MB, near enough 89% smaller than the last release"]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, security)","llm: publish; 4 citas verificadas, 1 rechazadas"]},"sources":[{"url":"https://github.com/open-webui/open-webui/releases/tag/v0.11.4","title":"open-webui/open-webui v0.11.4","retrieved_at":"2026-10-13T05:51:43Z","sha256":"74ff35906189e5533725f11c620064c210f1f17877a4ccb2d6ffeba03ccd25bc","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--open-webui--open-webui--v0.11.4","related":[],"org":{"slug":"open-webui","name":"Open WebUI","kind":"proyecto","url":"https://openwebui.com"},"tool":{"slug":"open-webui","name":"Open WebUI"},"version":"v0.11.4"},{"id":"hf:Qwen/Qwen-Image-2.1-PE-I2I","kind":"model_release","title":"Qwen publica Qwen-Image-2.1-PE-I2I","date":"2026-09-20","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"Qwen/Qwen-Image-2.1-PE-I2I","url":"https://huggingface.co/Qwen/Qwen-Image-2.1-PE-I2I","entity":"qwen","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 9.4 B parámetros, licencia other:qwen-research, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":9409813744,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"other:qwen-research","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"created_at","value":"2026-09-20T08:46:47.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"Qwen3_5ForConditionalGeneration","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/Qwen/Qwen-Image-2.1-PE-I2I","title":"Hugging Face · Qwen/Qwen-Image-2.1-PE-I2I","retrieved_at":"2026-10-13T08:54:49Z","sha256":"1aa4f53f9ea5c7c44f1307f19271a19d23a23c83c1fb7e516d8337b99e9b0a92","license":null},{"url":"https://huggingface.co/Qwen/Qwen-Image-2.1-PE-I2I/blob/main/config.json","title":"Qwen/Qwen-Image-2.1-PE-I2I · config.json","retrieved_at":"2026-10-13T08:54:49Z","sha256":"8b917c06bd646e9a8d9fef85e8ea58bca780caad20850a65007082f913e340d8","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"Qwen/Qwen-Image-2.1-PE-I2I","params_total":9409813744,"architecture":"Qwen3_5ForConditionalGeneration","context_length":262144,"license":"other:qwen-research","gated":false,"pipeline":null,"variants":[],"kv_bytes_per_token":32768,"quants":[]},"slug":"hf--qwen--qwen-image-2.1-pe-i2i","related":[],"org":{"slug":"qwen","name":"Qwen (Alibaba)","kind":"laboratorio","url":"https://qwen.ai"}},{"id":"rss:xai-news:ce5508c16c0a","kind":"announcement","title":"Introducing Grok Voice Transcribe 2.0 | SpaceXSI","date":"2026-09-18","observed_at":"2026-10-13T06:23:24Z","subject":{"type":"feed","ref":"xai-news","url":"https://x.ai/news/grok-voice-transcribe-2","entity":"xai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":[],"summary":{"why":"Se anuncia un nuevo modelo de speech-to-text (STT) que promete alta precisión y eficiencia de costes."},"highlights":[{"text":"Lanzamiento de Grok Voice Transcribe 2.0, un nuevo modelo de transcripción de voz.","quote":"Announcing SpaceXSI's newest speech-to-text model","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:23:24Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas","publish con severidad minor: queda en observación","corrección editorial: lanzamiento real nunca es minor (la severidad minor era un fallo de calibración del prompt)"]},"sources":[{"url":"https://x.ai/news/grok-voice-transcribe-2","title":"Introducing Grok Voice Transcribe 2.0 | SpaceXSI","retrieved_at":"2026-10-13T05:52:16Z","sha256":"a1b1465177bba81246ead3629f6f5a67af9984f4a8669e1c0bcc51609a1f6ce1","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--xai-news--ce5508c16c0a","related":[],"org":{"slug":"xai","name":"xAI","kind":"laboratorio","url":"https://x.ai"}},{"id":"github:NVIDIA/TensorRT-LLM:v1.3.0rc27","kind":"software_release","title":"TensorRT-LLM v1.3.0rc27","date":"2026-09-18","observed_at":"2026-10-13T06:13:15Z","subject":{"type":"repo","ref":"NVIDIA/TensorRT-LLM","url":"https://github.com/NVIDIA/TensorRT-LLM/releases/tag/v1.3.0rc27","entity":"nvidia","channel":"auto-build"},"status":"published","severity":"notable","action":"watch","tags":["api","breaking","new-model"],"affects":["tensorrt-llm"],"summary":{"why":"Esta versión incluye cambios disruptivos (breaking changes) en la API y soporte para nuevos modelos y arquitecturas (Blackwell/SM107), además de optimizaciones críticas para despliegues distribuidos."},"highlights":[{"text":"Cambios disruptivos en la API de especulación y runtime de C++","quote":"Remove deprecated two-model speculative decoding and related configuration fields (BREAKING)","source":0},{"text":"Soporte para nuevos modelos y capacidades multimodales","quote":"Add MiniMax-H3 VisualGen support for text-to-video with audio and first/last-frame-to-video with audio","source":0},{"text":"Optimización para hardware Blackwell (SM107) y kernels NVFP4","quote":"Add SM107 NVFP4 CuTe DSL fused MoE kernels and integration","source":0},{"text":"Nuevas APIs compatibles con Anthropic para el servidor","quote":"Add Anthropic-compatible Messages, token-counting, and Message Batches APIs to `trtllm-serve`","source":0},{"text":"Mejoras en el rendimiento de modelos Qwen y DeepSeek","quote":"Optimize Qwen3.5 and Qwen3.8 Flash-Next generation, MoE, GDN, and sampling","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:13:15Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, deprecation)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/NVIDIA/TensorRT-LLM/releases/tag/v1.3.0rc27","title":"NVIDIA/TensorRT-LLM v1.3.0rc27","retrieved_at":"2026-10-13T05:51:39Z","sha256":"6ba07424659d7bfe792965f8c087b2f73085d3faa7bfdce709a1b581f9204937","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--nvidia--tensorrt-llm--v1.3.0rc27","related":["hf:Qwen/Qwen3.8-Flash-Next"],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"},"tool":{"slug":"tensorrt-llm","name":"TensorRT-LLM"},"version":"v1.3.0rc27"},{"id":"github:sgl-project/sglang:v0.5.20","kind":"software_release","title":"sglang v0.5.20","date":"2026-09-18","observed_at":"2026-10-13T06:12:47Z","subject":{"type":"repo","ref":"sgl-project/sglang","url":"https://github.com/sgl-project/sglang/releases/tag/v0.5.20","entity":"sglang","channel":"stable"},"status":"published","severity":"notable","action":"try","tags":["performance","security"],"affects":["sglang"],"summary":{"why":"Esta actualización introduce soporte para nuevos modelos, optimizaciones críticas de rendimiento para DeepSeek-V4 en hardware Blackwell y RTX, y cambios en la gestión de memoria de la API que requieren atención."},"highlights":[{"text":"Soporte para nuevos modelos autoregresivos y de difusión.","quote":"New models in this release","source":0},{"text":"Optimización masiva de DeepSeek-V4 en arquitecturas Blackwell (B200).","quote":"TRT-LLM attention kernels now cover DeepSeek-V4's CSA and HCA layers on SM100 and SM103: about 1.2x faster prefill and 1.45x faster decode than FlashMLA at the kernel level on B200","source":0},{"text":"Mejora de rendimiento para DeepSeek-V4 en GPUs RTX PRO 6000.","quote":"On 4x RTX PRO 6000, DeepSeek-V4-Flash decode TPOT drops from 36.1 to 10.5 ms at batch 1","source":0},{"text":"Cambio en el comportamiento de la API de respuestas (ahora es opt-in).","quote":"Responses API storage is opt-in. /v1/responses no longer retains results in memory unless the server starts with --enable-response-store.","source":0},{"text":"Retirada del soporte para CUDA 12.","quote":"The CUDA 12 lane is retired; v0.5.19 was the last release with -cu12x wheels and images","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:12:47Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, deprecation, feature, performance)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/sgl-project/sglang/releases/tag/v0.5.20","title":"sgl-project/sglang v0.5.20","retrieved_at":"2026-10-13T05:51:38Z","sha256":"99d5048b4992ef09c17613468392ca873a26d4630d009f654bb7bf2d9b3cfacc","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--sgl-project--sglang--v0.5.20","related":[],"org":{"slug":"sglang","name":"SGLang","kind":"proyecto","url":"https://sglang.ai"},"tool":{"slug":"sglang","name":"SGLang"},"version":"v0.5.20"},{"id":"github:anthropics/claude-code:v2.1.276","kind":"software_release","title":"claude-code v2.1.276","date":"2026-09-18","observed_at":"2026-10-13T05:58:16Z","subject":{"type":"repo","ref":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.276","entity":"anthropic","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["breaking","fix"],"affects":["claude-code"],"summary":{"why":"Corrige una regresión crítica en la versión anterior que impedía el uso de la herramienta cuando se utiliza un proxy o gateway mediante la variable ANTHROPIC_BASE_URL."},"highlights":[{"text":"Corrección de error 400 al usar proxies o gateways con ANTHROPIC_BASE_URL","quote":"Fixed every request failing with `400 … Input tag 'advisor_20260301'` when `ANTHROPIC_BASE_URL` points at a proxy or gateway","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:58:16Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: drop (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.276","title":"anthropics/claude-code v2.1.276","retrieved_at":"2026-10-13T05:51:19Z","sha256":"4cb3a679bc1546669684d0e7b66cf7794ca568aa755815f1dde3b0ead2928132","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--anthropics--claude-code--v2.1.276","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"},"tool":{"slug":"claude-code","name":"Claude Code"},"version":"v2.1.276"},{"id":"hf:microsoft/FrogNano-4B-2609","kind":"model_release","title":"microsoft publica FrogNano-4B-2609","date":"2026-09-17","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"microsoft/FrogNano-4B-2609","url":"https://huggingface.co/microsoft/FrogNano-4B-2609","entity":"microsoft","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Modelo nuevo: 4.7 B parámetros, licencia mit, 22 cuantizaciones GGUF disponibles."},"highlights":[],"facts":[{"field":"params_total","value":4659865088,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"mit","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"created_at","value":"2026-09-17T00:13:57.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"Qwen3_5ForConditionalGeneration","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}},{"field":"gguf_quants","value":22,"evidence":{"source":2,"locator":"siblings[].size"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/microsoft/FrogNano-4B-2609","title":"Hugging Face · microsoft/FrogNano-4B-2609","retrieved_at":"2026-10-13T08:54:55Z","sha256":"7c6b0573d736b3586b53344e3da1363484684d99ec87c6617d53562e4a87af0d","license":null},{"url":"https://huggingface.co/microsoft/FrogNano-4B-2609/blob/main/config.json","title":"microsoft/FrogNano-4B-2609 · config.json","retrieved_at":"2026-10-13T08:54:56Z","sha256":"e699c9d152b18ff2ad9ba09bc35867a9949a0431f1f38f171ee2de470795e397","license":null},{"url":"https://huggingface.co/bartowski/FrogNano-4B-2609-GGUF","title":"Hugging Face · bartowski/FrogNano-4B-2609-GGUF","retrieved_at":"2026-10-13T08:54:56Z","sha256":"54aa1c77f580f0dfd8b6e47d6c6d2809855eeba5059cdb2f1ab55026ef433ac3","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"microsoft/FrogNano-4B-2609","params_total":4659865088,"architecture":"Qwen3_5ForConditionalGeneration","context_length":262144,"license":"mit","gated":false,"pipeline":null,"variants":[],"kv_bytes_per_token":32768,"quants":[{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"IQ2_M","bytes":1653872416},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q2_K","bytes":1797693216},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"IQ3_XXS","bytes":1975582496},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q3_K_S","bytes":2035466016},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"IQ3_XS","bytes":2050211616},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q3_K_M","bytes":2140753696},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q3_K_L","bytes":2244648736},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"IQ3_M","bytes":2345881376},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"IQ4_XS","bytes":2506567456},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q4_K_S","bytes":2615664416},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q4_0","bytes":2715739936},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"IQ4_NL","bytes":2787061536},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q4_K_M","bytes":2797526816},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q4_1","bytes":2919556896},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q4_K_L","bytes":2969456416},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q5_K_S","bytes":3154749216},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q5_K_M","bytes":3360941856},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q6_K_S","bytes":3611739936},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q6_K","bytes":3757762336},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q6_K_L","bytes":3918386976},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"Q8_0","bytes":4622130976},{"repo":"bartowski/FrogNano-4B-2609-GGUF","quant":"BF16","bytes":8665619936}]},"slug":"hf--microsoft--frognano-4b-2609","related":[],"org":{"slug":"microsoft","name":"Microsoft","kind":"empresa","url":"https://www.microsoft.com"}},{"id":"github:openai/codex:rust-v0.155.0","kind":"software_release","title":"codex 0.155.0","date":"2026-09-17","observed_at":"2026-10-13T06:09:21Z","subject":{"type":"repo","ref":"openai/codex","url":"https://github.com/openai/codex/releases/tag/rust-v0.155.0","entity":"openai","channel":"stable"},"status":"published","severity":"notable","action":"try","tags":[],"affects":["codex"],"summary":{"why":"Esta actualización introduce capacidades experimentales de voz en tiempo real y mejoras significativas en la seguridad de credenciales y sandboxing, además de soporte para AWS Bedrock."},"highlights":[{"text":"Conversaciones experimentales con transcripción en vivo y control de micrófono.","quote":"Added experimental `/voice` conversations with live transcripts and microphone controls on supported builds, enabled through `/experimental`.","source":0},{"text":"Soporte para obtener credenciales de AWS desde comandos configurados en Amazon Bedrock.","quote":"Amazon Bedrock can now obtain AWS credentials from configured commands, with caching, expiration-based refresh, and authentication recovery.","source":0},{"text":"Mejoras de seguridad para evitar escapes de procesos de Windows desde sandboxes WSL.","quote":"Blocked Windows-process escapes from restricted WSL sandboxes and hardened brokered shell snapshots against credential exposure.","source":0},{"text":"Verificación de Touch ID para solicitudes MCP en sesiones TUI locales en macOS.","quote":"Added Touch ID verification for MCP requests in local TUI sessions on supported Macs.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:09:21Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (feature)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/openai/codex/releases/tag/rust-v0.155.0","title":"openai/codex 0.155.0","retrieved_at":"2026-10-13T05:51:31Z","sha256":"54066c9682b5f66eb8420f047313b6a1b5d79a9ae18aaee0d8e5dd440e78e410","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--openai--codex--rust-v0.155.0","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"},"tool":{"slug":"codex","name":"Codex CLI"},"version":"rust-v0.155.0"},{"id":"hf:inclusionAI/Realtime-Venus","kind":"model_release","title":"inclusionAI publica Realtime-Venus","date":"2026-09-16","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"inclusionAI/Realtime-Venus","url":"https://huggingface.co/inclusionAI/Realtime-Venus","entity":"inclusionai","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: licencia apache-2.0, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"license","value":"apache-2.0","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"any-to-any","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-16T08:02:28.000Z","evidence":{"source":0,"locator":"createdAt"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/inclusionAI/Realtime-Venus","title":"Hugging Face · inclusionAI/Realtime-Venus","retrieved_at":"2026-10-13T08:55:15Z","sha256":"180d924d973af316d06649cef4a2cff2e2146baa96c181d470fff78c3d9f9f5b","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"inclusionAI/Realtime-Venus","params_total":null,"architecture":null,"context_length":null,"license":"apache-2.0","gated":false,"pipeline":"any-to-any","variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--inclusionai--realtime-venus","related":[],"org":{"slug":"inclusionai","name":"inclusionAI (Ant Group)","kind":"laboratorio","url":"https://github.com/inclusionAI"}},{"id":"rss:cohere-changelog:770f012e891d","kind":"announcement","title":"Announcing Cohere's North Small Translate | Cohere","date":"2026-09-16","observed_at":"2026-10-13T06:23:42Z","subject":{"type":"feed","ref":"cohere-changelog","url":"https://docs.cohere.com/changelog/north-small-translate-1-0","entity":"cohere","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":[],"summary":{"why":"Cohere lanza un nuevo modelo especializado en traducción (North Small Translate), lo que amplía las opciones de modelos compactos para tareas específicas de lenguaje."},"highlights":[{"text":"Lanzamiento de North Small Translate de Cohere","quote":"the release of North Small Translate, Cohere","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:23:42Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas","publish con severidad minor: queda en observación","corrección editorial: lanzamiento real nunca es minor (la severidad minor era un fallo de calibración del prompt)"]},"sources":[{"url":"https://docs.cohere.com/changelog/north-small-translate-1-0","title":"Announcing Cohere's North Small Translate | Cohere","retrieved_at":"2026-10-13T05:52:20Z","sha256":"b6fc11f4849961af6b7f27048a2d14e60b471411aaf63db8dfabd17155270988","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--cohere-changelog--770f012e891d","related":[],"org":{"slug":"cohere","name":"Cohere","kind":"laboratorio","url":"https://cohere.com"}},{"id":"github:openvinotoolkit/openvino:2026.4.0","kind":"software_release","title":"openvino 2026.4.0","date":"2026-09-16","observed_at":"2026-10-13T06:14:37Z","subject":{"type":"repo","ref":"openvinotoolkit/openvino","url":"https://github.com/openvinotoolkit/openvino/releases/tag/2026.4.0","entity":"intel","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["breaking","new-model","performance"],"affects":["openvino"],"summary":{"why":"Esta actualización introduce soporte para modelos clave como Gemma 4 y Qwen3, nuevas técnicas de aceleración (MTP, DFlash) y elimina namespaces y requisitos de hardware (AVX2), lo que requiere atención inmediata para evitar roturas en pipelines existentes."},"highlights":[{"text":"Soporte para nuevos modelos de GenAI en CPU, GPU y NPU (Gemma 4, Qwen3, etc.)","quote":"New models supported: On CPU: Gemma-3n On CPU, GPU: Kokoro-82M, Qwen3-VL-4B with EAGLE-3, Qwen3-ASR, Muse Glimmer 30B, Qwen3.8 27B, Gemma 4 12B, Hy-MT2-1.8B, DeepSeek OCR-2, and Granite 4.0 H Micro","source":0},{"text":"Implementación de Multi-Token Prediction (MTP) para mayor rendimiento en Gemma 4 y Qwen3","quote":"OpenVINO™ GenAI adds Multi-Token Prediction (MTP) speculative decoding for Gemma 4, Qwen3.5, and Qwen3.6 on CPUs and GPUs, enabling higher throughput and lower latency without sacrificing accuracy.","source":0},{"text":"Eliminación del namespace 'openvino.runtime' y cambios en la API","quote":"The deprecated `openvino.runtime` namespace has been removed. Please use the `openvino` namespace directly.","source":0},{"text":"Requisito mínimo de hardware: soporte obligatorio para instrucciones AVX2","quote":"CPU plugin now requires support for the AVX2 instruction set as a minimum system requirement. The SSE instruction set will no longer be supported.","source":0},{"text":"Soporte para ASR en Node.js mediante ASRPipeline","quote":"OpenVINO™ GenAI introduces support for ASRPipeline in Node.js, enabling JavaScript developers to run automatic speech recognition (such as Whisper and Qwen3-ASR)","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:14:37Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (deprecation, feature, new-architecture)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/openvinotoolkit/openvino/releases/tag/2026.4.0","title":"openvinotoolkit/openvino 2026.4.0","retrieved_at":"2026-10-13T05:51:46Z","sha256":"fcd5097b4ae25ec78965d20670030014fac985a63c8f508901bb58b674e75632","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--openvinotoolkit--openvino--2026.4.0","related":[],"org":{"slug":"intel","name":"Intel","kind":"empresa","url":"https://www.intel.com"},"tool":{"slug":"openvino","name":"OpenVINO"},"version":"2026.4.0"},{"id":"hf:nvidia/RT-DETR-Hand-Detector-v1","kind":"model_release","title":"nvidia publica RT-DETR-Hand-Detector-v1","date":"2026-09-15","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"nvidia/RT-DETR-Hand-Detector-v1","url":"https://huggingface.co/nvidia/RT-DETR-Hand-Detector-v1","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: licencia other:nvidia-software-and-model-evaluation-license, acceso restringido (gated)."},"highlights":[],"facts":[{"field":"license","value":"other:nvidia-software-and-model-evaluation-license","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":true,"evidence":{"source":0,"locator":"gated"}},{"field":"created_at","value":"2026-09-15T16:58:11.000Z","evidence":{"source":0,"locator":"createdAt"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/nvidia/RT-DETR-Hand-Detector-v1","title":"Hugging Face · nvidia/RT-DETR-Hand-Detector-v1","retrieved_at":"2026-10-13T08:55:01Z","sha256":"d87fa93b6b40ce07a5ac8dbfd84d96a7d36791d9c96c2da45f5e11f70f354d39","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"nvidia/RT-DETR-Hand-Detector-v1","params_total":null,"architecture":null,"context_length":null,"license":"other:nvidia-software-and-model-evaluation-license","gated":true,"pipeline":null,"variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--nvidia--rt-detr-hand-detector-v1","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}},{"id":"github:LostRuins/koboldcpp:v1.121","kind":"software_release","title":"koboldcpp-1.121","date":"2026-09-15","observed_at":"2026-10-13T06:14:26Z","subject":{"type":"repo","ref":"LostRuins/koboldcpp","url":"https://github.com/LostRuins/koboldcpp/releases/tag/v1.121","entity":"koboldcpp","channel":"stable"},"status":"published","severity":"notable","action":"try","tags":["feature","new-architecture","new-model"],"affects":["koboldcpp"],"summary":{"why":"Esta versión introduce soporte para multimedia (audio/imágenes) en modelos Minimax H3, soporte para video LoRAs y una nueva funcionalidad de offload parcial de capas similar a llama.cpp."},"highlights":[{"text":"Soporte de referencias multimedia para Minimax H3","quote":"NEW: Media reference support added for Minimax H3!","source":0},{"text":"Soporte para video LoRAs","quote":"Added support for video LoRAs (thanks @wbruna)","source":0},{"text":"Nueva bandera para offload parcial de capas (similar a llama.cpp)","quote":"Added a flag `--ffncpu` which behaves like llama.cpp `--n-cpu-ffn`, an alternative way to partial offload layers","source":0},{"text":"Selector de LoRA en tiempo de ejecución para Stable-UI","quote":"Runtime LoRA selector in included stable-ui (http://localhost:5001/sdui)","source":0},{"text":"Análisis de metadatos de archivos .safetensors","quote":"`--analyze` feature can now analyze `.safetensor` file metadata in addition to `.gguf` files.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:14:26Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (feature)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/LostRuins/koboldcpp/releases/tag/v1.121","title":"LostRuins/koboldcpp koboldcpp-1.121","retrieved_at":"2026-10-13T05:51:44Z","sha256":"46d71eadf27275567887f4885f557ee8727dc1101d6d418625c5a8fc7f236ab3","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--lostruins--koboldcpp--v1.121","related":[],"org":{"slug":"koboldcpp","name":"KoboldCpp","kind":"proyecto","url":"https://koboldai.org"},"tool":{"slug":"koboldcpp","name":"KoboldCpp"},"version":"v1.121"},{"id":"github:anthropics/claude-code:v2.1.273","kind":"software_release","title":"claude-code v2.1.273","date":"2026-09-15","observed_at":"2026-10-13T05:58:55Z","subject":{"type":"repo","ref":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.273","entity":"anthropic","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["feature","performance","security"],"affects":["claude-code"],"summary":{"why":"Esta actualización incluye mejoras críticas en la gestión de permisos de Bash, optimización del uso de contexto (context window) y cambios en el comportamiento de los clasificadores para entornos de servidor (Bedrock/Vertex)."},"highlights":[{"text":"Nuevos headers para gateways de LLM y soporte de optimización de contexto.","quote":"Added `x-claude-code-request-class`, `x-claude-code-agent-type`, `x-claude-code-prev-tool-durations`, `x-claude-code-compaction` and `x-claude-code-context-compacted` request headers for LLM gateways","source":0},{"text":"Corrección de vulnerabilidad en el checker de permisos de Bash.","quote":"Fixed Bash commands the permission checker cannot fully analyze skipping the prompt under `permissions.blockReadsOutsideWorkingDirectories`, and a subshell hiding a dangerous `rm` in bypass mode","source":0},{"text":"Mejora en la precisión del medidor de contexto para evitar disparos prematuros de compactación.","quote":"Fixed the context meter and auto-compact counting advisor-tool turns at roughly twice their real context size","source":0},{"text":"Cambio en el uso de clasificadores para Bedrock, Vertex y Foundry.","quote":"Changed auto mode on Bedrock, Vertex and Foundry to use the local classifier by default for now","source":0},{"text":"Mejoras de rendimiento en sesiones largas.","quote":"Improved responsiveness in long sessions: hook progress and sub-agent activity no longer re-process the whole conversation on every update","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:58:55Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (feature)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.273","title":"anthropics/claude-code v2.1.273","retrieved_at":"2026-10-13T05:51:19Z","sha256":"394997b01870f14ddfbf6ca823da29412f58382bd381f03c72b868091555d5d6","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--anthropics--claude-code--v2.1.273","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"},"tool":{"slug":"claude-code","name":"Claude Code"},"version":"v2.1.273"},{"id":"hf:google/embeddinggemma-2","kind":"model_release","title":"google publica embeddinggemma-2","date":"2026-09-14","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"google/embeddinggemma-2","url":"https://huggingface.co/google/embeddinggemma-2","entity":"google","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Modelo nuevo: 0.7 B parámetros, licencia apache-2.0, 6 cuantizaciones GGUF disponibles."},"highlights":[],"facts":[{"field":"params_total","value":744371512,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"apache-2.0","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"feature-extraction","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-14T14:53:42.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"EmbeddingGemma2Model","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}},{"field":"gguf_quants","value":6,"evidence":{"source":2,"locator":"siblings[].size"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/google/embeddinggemma-2","title":"Hugging Face · google/embeddinggemma-2","retrieved_at":"2026-10-13T08:54:52Z","sha256":"25787888e7ce6da7af49df627c76316da128449464664b32083f843c184c6fc6","license":null},{"url":"https://huggingface.co/google/embeddinggemma-2/blob/main/config.json","title":"google/embeddinggemma-2 · config.json","retrieved_at":"2026-10-13T08:54:52Z","sha256":"26776d34e57dd4e360ac9b6e1f59c9c58e9272d5a5205048edb4c0f7d14d93da","license":null},{"url":"https://huggingface.co/unsloth/embeddinggemma-2-GGUF","title":"Hugging Face · unsloth/embeddinggemma-2-GGUF","retrieved_at":"2026-10-13T08:54:53Z","sha256":"fdf642d309929160948bdcbb09786935120f9477e07c5bbd369a3ff8b84d5876","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"google/embeddinggemma-2","params_total":744371512,"architecture":"EmbeddingGemma2Model","context_length":262144,"license":"apache-2.0","gated":false,"pipeline":"feature-extraction","variants":[],"kv_bytes_per_token":49152,"quants":[{"repo":"unsloth/embeddinggemma-2-GGUF","quant":"Q4_K_XL","bytes":175673856},{"repo":"unsloth/embeddinggemma-2-GGUF","quant":"Q5_K_XL","bytes":210055680},{"repo":"unsloth/embeddinggemma-2-GGUF","quant":"Q6_K_XL","bytes":248812032},{"repo":"unsloth/embeddinggemma-2-GGUF","quant":"Q8_0","bytes":309855520},{"repo":"unsloth/embeddinggemma-2-GGUF","quant":"BF16","bytes":557950240},{"repo":"unsloth/embeddinggemma-2-GGUF","quant":"F16","bytes":557950240}]},"slug":"hf--google--embeddinggemma-2","related":["github:huggingface/transformers:v5.19.0","github:ollama/ollama:v0.40.4-rc0","github:ollama/ollama:v0.40.4-rc1","rss:deepmind:c11a83f73b42"],"org":{"slug":"google","name":"Google","kind":"empresa","url":"https://deepmind.google"}},{"id":"github:turboderp-org/exllamav3:v1.5.0","kind":"software_release","title":"exllamav3 1.5.0","date":"2026-09-13","observed_at":"2026-10-13T06:13:47Z","subject":{"type":"repo","ref":"turboderp-org/exllamav3","url":"https://github.com/turboderp-org/exllamav3/releases/tag/v1.5.0","entity":"exllama","channel":"stable"},"status":"published","severity":"notable","action":"try","tags":["new-model","performance"],"affects":["exllamav3"],"summary":{"why":"Esta actualización introduce mejoras significativas de rendimiento en prefill y decode, especialmente para modelos MoE y arquitecturas específicas, además de un nuevo modo experimental de zero-copy para CPU offload en Linux."},"highlights":[{"text":"Soporte para Nemotron-3-Super","quote":"Fix NemotronHForCausalLM (support Nemotron-3-Super)","source":0},{"text":"Modo experimental de zero-copy pinned arena para CPU offload (solo Linux)","quote":"Experimental zero-copy pinned arena mode for CPU offload (Linux only for now)","source":0},{"text":"Mejoras de velocidad en cuantización y modelos MoE","quote":"Faster quantization","source":0},{"text":"Mejoras en el rendimiento de prefill y decode para modelos MoE","quote":"Faster MoE prefill and decode (see below)","source":0},{"text":"Mejoras en el speculative decoding para modelos MoE","quote":"Improved speculative decoding for MoE models","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:13:47Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: drop (sin señales)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/turboderp-org/exllamav3/releases/tag/v1.5.0","title":"turboderp-org/exllamav3 1.5.0","retrieved_at":"2026-10-13T05:51:42Z","sha256":"120ada2ebeb00db5d80b9c11c2d7289f96920dde0567be79fc31018f62292255","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--turboderp-org--exllamav3--v1.5.0","related":[],"org":{"slug":"exllama","name":"ExLlama","kind":"proyecto","url":"https://github.com/turboderp-org/exllamav3"},"tool":{"slug":"exllamav3","name":"ExLlamaV3"},"version":"v1.5.0"},{"id":"rss:cohere-blog:ce6b8a6b961e","kind":"announcement","title":"Introducing North Small Translate | Cohere","date":"2026-09-10","observed_at":"2026-10-13T06:23:57Z","subject":{"type":"feed","ref":"cohere-blog","url":"https://cohere.com/blog/north-small-translate","entity":"cohere","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":[],"summary":{"why":"Cohere lanza un nuevo modelo especializado en traducción diseñado para ser eficiente en velocidad y costes, ideal para implementaciones que requieren baja latencia."},"highlights":[{"text":"Modelo optimizado para rendimiento y eficiencia de costes en traducción.","quote":"Outsized performance, right-sized footprint — translation built for speed and cost efficiency.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:23:57Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas","publish con severidad minor: queda en observación","corrección editorial: lanzamiento real nunca es minor (la severidad minor era un fallo de calibración del prompt)"]},"sources":[{"url":"https://cohere.com/blog/north-small-translate","title":"Introducing North Small Translate | Cohere","retrieved_at":"2026-10-13T05:52:24Z","sha256":"4793f63f4b2a34366b1712a1bdfed3a42205856412c3edd1435b0ebcfcdb5ae8","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--cohere-blog--ce6b8a6b961e","related":[],"org":{"slug":"cohere","name":"Cohere","kind":"laboratorio","url":"https://cohere.com"}},{"id":"rss:openai:b32e9b847135","kind":"announcement","title":"Introducing the Agents API","date":"2026-09-10","observed_at":"2026-10-13T06:16:58Z","subject":{"type":"feed","ref":"openai","url":"https://openai.com/index/introducing-the-agents-api","entity":"openai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["api","new-model"],"affects":[],"summary":{"why":"OpenAI lanza una nueva API gestionada para la creación y despliegue de agentes autónomos con capacidades de orquestación y uso de herramientas."},"highlights":[{"text":"Lanzamiento de la Agents API para construir y ejecutar agentes en la nube.","quote":"Build and launch cloud agents with the Agents API","source":0},{"text":"Servicio gestionado que utiliza el harness de Codex para la orquestación y sesiones de larga duración.","quote":"a managed service powered by the Codex harness for orchestration, long-running sessions, and tool use.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:16:58Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 2 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://openai.com/index/introducing-the-agents-api","title":"Introducing the Agents API","retrieved_at":"2026-10-13T05:51:53Z","sha256":"0d4c07fdbb755f2eb37b5e92d311a31d3cd7017047a7f6a8816a01f4b7ceda9d","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--openai--b32e9b847135","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"}},{"id":"rss:openai:b8a75fb67d81","kind":"announcement","title":"Build more natural voice experiences with GPT‑Live‑1 in the API","date":"2026-09-10","observed_at":"2026-10-13T06:16:55Z","subject":{"type":"feed","ref":"openai","url":"https://openai.com/index/introducing-gpt-live-1-in-the-api","entity":"openai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["api","new-model"],"affects":[],"summary":{"why":"OpenAI introduce un nuevo modelo para conversaciones de voz full-duplex con mayor capacidad de seguimiento de instrucciones y soporte para telefonía vía API."},"highlights":[{"text":"Introducción de GPT-Live-1 para experiencias de voz naturales y full-duplex.","quote":"GPT‑Live‑1 brings natural, full-duplex voice conversations to the API","source":0},{"text":"Mejoras en el seguimiento de instrucciones y voces personalizadas.","quote":"with stronger instruction following, custom voices, and telephony support","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:16:55Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 2 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://openai.com/index/introducing-gpt-live-1-in-the-api","title":"Build more natural voice experiences with GPT‑Live‑1 in the API","retrieved_at":"2026-10-13T05:51:53Z","sha256":"40cb6ee92f55196c7ede6e57295704aff8984be98c2073325b79a3ce7ab838b0","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--openai--b8a75fb67d81","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"}},{"id":"hf:allenai/AstaBrief_8B_SFT","kind":"model_release","title":"allenai publica AstaBrief_8B_SFT","date":"2026-09-10","observed_at":"2026-10-13T06:14:49Z","subject":{"type":"model","ref":"allenai/AstaBrief_8B_SFT","url":"https://huggingface.co/allenai/AstaBrief_8B_SFT","entity":"allenai","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: licencia apache-2.0, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"license","value":"apache-2.0","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"text-generation","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-10T23:36:30.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"Qwen3ForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":40960,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T06:14:49Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/allenai/AstaBrief_8B_SFT","title":"Hugging Face · allenai/AstaBrief_8B_SFT","retrieved_at":"2026-10-13T05:51:53Z","sha256":"bd4ae9986ff6415a85c62d0d7f8e59878da8c206808ca09bdfeba86c8c9e329f","license":null},{"url":"https://huggingface.co/allenai/AstaBrief_8B_SFT/blob/main/config.json","title":"allenai/AstaBrief_8B_SFT · config.json","retrieved_at":"2026-10-13T05:51:53Z","sha256":"a7e85611ee66c8f227a761f95e9e6acbd79f884266d3ef353f9a67bd1030f5c8","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"allenai/AstaBrief_8B_SFT","params_total":null,"architecture":"Qwen3ForCausalLM","context_length":40960,"license":"apache-2.0","gated":false,"pipeline":"text-generation","variants":[],"kv_bytes_per_token":147456,"quants":[]},"slug":"hf--allenai--astabrief_8b_sft","related":[],"org":{"slug":"allenai","name":"Allen Institute for AI","kind":"laboratorio","url":"https://allenai.org"}},{"id":"hf:deepseek-ai/DeepSeek-V4.1-Flash","kind":"model_release","title":"deepseek-ai publica DeepSeek-V4.1-Flash","date":"2026-09-10","observed_at":"2026-10-13T06:14:49Z","subject":{"type":"model","ref":"deepseek-ai/DeepSeek-V4.1-Flash","url":"https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash","entity":"deepseek-ai","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 763.2 B parámetros, licencia mit, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":763205315794,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"mit","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"image-text-to-text","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-10T02:17:58.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"DeepseekV41ForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":1048576,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T06:14:49Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash","title":"Hugging Face · deepseek-ai/DeepSeek-V4.1-Flash","retrieved_at":"2026-10-13T05:51:47Z","sha256":"da5f504ceaf7ba76df477a0c48f1584a4dad037ea6a237609e67ccad983ed6cb","license":null},{"url":"https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash/blob/main/config.json","title":"deepseek-ai/DeepSeek-V4.1-Flash · config.json","retrieved_at":"2026-10-13T05:51:47Z","sha256":"0110b6f9349cf6404bef91bfe8a491be248d3d5a044592d3c443dc9073fa0096","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"deepseek-ai/DeepSeek-V4.1-Flash","params_total":763205315794,"architecture":"DeepseekV41ForCausalLM","context_length":1048576,"license":"mit","gated":false,"pipeline":"image-text-to-text","variants":[],"kv_bytes_per_token":81920,"quants":[]},"slug":"hf--deepseek-ai--deepseek-v4.1-flash","related":["github:sgl-project/sglang:v0.5.21"],"org":{"slug":"deepseek","name":"DeepSeek","kind":"laboratorio","url":"https://www.deepseek.com"}},{"id":"github:turboderp-org/exllamav3:v1.4.9","kind":"software_release","title":"exllamav3 1.4.9","date":"2026-09-10","observed_at":"2026-10-13T06:13:51Z","subject":{"type":"repo","ref":"turboderp-org/exllamav3","url":"https://github.com/turboderp-org/exllamav3/releases/tag/v1.4.9","entity":"exllama","channel":"stable"},"status":"published","severity":"notable","action":"try","tags":["new-model","performance"],"affects":["exllamav3"],"summary":{"why":"Introduce soporte para nuevos modelos (DeepseekV4, LFM2.5, Spark2.5) y optimizaciones críticas en el offloading a CPU y tiempos de importación."},"highlights":[{"text":"Soporte para la torre de visión de DeepseekV4","quote":"Add DeepseekV4ForCausalLM vision tower (supports DeepseekV4-Flash-Vision-Exp)","source":0},{"text":"Soporte para modelos LFM2.5, Spark-X2.5 y GLM4.7-Flash","quote":"Add Lfm2ForCausalLM (LFM2.5 dense variant)\n- Add Spark2_5ForCausalLM (Spark-X2.5)\n- Add Glm4MoeLiteForCausalLM (GLM4.7-Flash)","source":0},{"text":"Mejoras en el offloading a CPU y soporte Intel AVX512BW","quote":"Faster CPU offloading, better Intel support (AVX512BW tier)","source":0},{"text":"Reducción significativa en el tiempo de importación","quote":"Greatly reduced import time, vendor in subset of flash-linear-attention to avoid initializing whole library plus Transformers","source":0},{"text":"Corrección de fuga de VRAM al descargar la torre de visión","quote":"Fix VRAM leak after vision tower offloading","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:13:51Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: drop (sin señales)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/turboderp-org/exllamav3/releases/tag/v1.4.9","title":"turboderp-org/exllamav3 1.4.9","retrieved_at":"2026-10-13T05:51:42Z","sha256":"0c2e32473890dfe624c1f15e7a31a71457d3145de8d57b4fb10e44c15dfea78e","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--turboderp-org--exllamav3--v1.4.9","related":[],"org":{"slug":"exllama","name":"ExLlama","kind":"proyecto","url":"https://github.com/turboderp-org/exllamav3"},"tool":{"slug":"exllamav3","name":"ExLlamaV3"},"version":"v1.4.9"},{"id":"github:anthropics/claude-code:v2.1.268","kind":"software_release","title":"claude-code v2.1.268","date":"2026-09-10","observed_at":"2026-10-13T05:59:25Z","subject":{"type":"repo","ref":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.268","entity":"anthropic","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["fix"],"affects":["claude-code"],"summary":{"why":"Esta actualización corrige un error crítico que causaba fallos en cada turno (HTTP 400) al usar endpoints compatibles con Anthropic (como proxies o modelos locales) y mejora la estabilidad de las sesiones con prompt caching y thinking."},"highlights":[{"text":"Corrección de errores HTTP 400 en endpoints de terceros compatibles con Anthropic.","quote":"Fixed every turn failing with HTTP 400 on third-party Anthropic-compatible endpoints (`ANTHROPIC_BASE_URL`) since 2.1.265","source":0},{"text":"Solución a la rotura de prompt caching y extended thinking en sesiones SDK.","quote":"Fixed prompt caching and extended thinking breaking mid-session for SDK sessions using `excludeDynamicSections`","source":0},{"text":"Mejora en el uso de CPU para sesiones inactivas de larga duración.","quote":"Fixed sustained high CPU usage: a busy loop in long-running idle sessions no longer pins a CPU core","source":0},{"text":"Corrección de errores de seguridad al mostrar tokens o contraseñas en errores de plugins.","quote":"Fixed plugin and marketplace errors showing a token or password from a git source URL","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:59:25Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, feature)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.268","title":"anthropics/claude-code v2.1.268","retrieved_at":"2026-10-13T05:51:19Z","sha256":"7fbba703f1faf975d15726d4c0c07f5a270d28560451f9576591e86f735aca9b","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--anthropics--claude-code--v2.1.268","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"},"tool":{"slug":"claude-code","name":"Claude Code"},"version":"v2.1.268"},{"id":"rss:nvidia-dev:fc4203ac554d","kind":"announcement","title":"CUDA Toolkit 13.4 Adds Windows on Arm Support and Greater Control over Shared GPUs","date":"2026-09-09","observed_at":"2026-10-13T06:20:06Z","subject":{"type":"feed","ref":"nvidia-dev","url":"https://developer.nvidia.com/blog/cuda-toolkit-13-4-adds-windows-on-arm-support-and-greater-control-over-shared-gpus/","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-architecture","performance"],"affects":["nvidia-driver"],"summary":{"why":"La introducción de soporte para Windows on Arm y un mejor control de GPUs compartidas es un cambio de infraestructura clave para el despliegue de modelos en hardware de nueva arquitectura."},"highlights":[{"text":"Soporte para Windows on Arm","quote":"CUDA Toolkit 13.4 Adds Windows on Arm Support","source":0},{"text":"Mayor control sobre GPUs compartidas","quote":"Greater Control over Shared GPUs","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:20:06Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 2 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://developer.nvidia.com/blog/cuda-toolkit-13-4-adds-windows-on-arm-support-and-greater-control-over-shared-gpus/","title":"CUDA Toolkit 13.4 Adds Windows on Arm Support and Greater Control over Shared GPUs","retrieved_at":"2026-10-13T05:51:54Z","sha256":"14d04ac643b5115c305204988bb5a694553ed1224997327c7da978f3e8d78b39","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--nvidia-dev--fc4203ac554d","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}},{"id":"rss:openai:333ccaae8148","kind":"announcement","title":"GPT-6 Astra: The next generation in intelligence for work","date":"2026-09-09","observed_at":"2026-10-13T06:17:02Z","subject":{"type":"feed","ref":"openai","url":"https://openai.com/index/gpt-6-astra-next-generation-work","entity":"openai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":[],"summary":{"why":"OpenAI anuncia su nueva generación de modelos con capacidades avanzadas de razonamiento y uso de computadora, lo que redefine el estado del arte para tareas de productividad."},"highlights":[{"text":"Capacidades avanzadas de razonamiento y uso de computadora.","quote":"advanced reasoning, computer use","source":0},{"text":"Mejor juicio en escritura y diseño.","quote":"stronger writing and design judgment","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:17:02Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 2 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://openai.com/index/gpt-6-astra-next-generation-work","title":"GPT-6 Astra: The next generation in intelligence for work","retrieved_at":"2026-10-13T05:51:53Z","sha256":"1564b3a3cf9bf1d49a065fca6ab8431d9847a935466feb3a025fc6c75a235a52","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--openai--333ccaae8148","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"}},{"id":"github:NVIDIA/TensorRT-LLM:v1.3.0rc26","kind":"software_release","title":"TensorRT-LLM v1.3.0rc26","date":"2026-09-09","observed_at":"2026-10-13T06:13:21Z","subject":{"type":"repo","ref":"NVIDIA/TensorRT-LLM","url":"https://github.com/NVIDIA/TensorRT-LLM/releases/tag/v1.3.0rc26","entity":"nvidia","channel":"auto-build"},"status":"published","severity":"notable","action":"watch","tags":["breaking","new-model"],"affects":["tensorrt-llm"],"summary":{"why":"Esta versión incluye cambios de ruptura (breaking changes) en la API y soporte para nuevos modelos críticos como DeepSeek-V4 y Qwen3, además de advertencias de compatibilidad con hardware Blackwell."},"highlights":[{"text":"Cambios de ruptura en la API (eliminación de TRTLLMSampler y rediseño de VisualGen).","quote":"Remove TRTLLMSampler (BREAKING)","source":0},{"text":"Soporte para nuevos modelos: DeepSeek-V4, Qwen3.8-Flash-Next y Kimi K3.","quote":"Add DeepSeek-V4 Hopper support #16940","source":0},{"text":"Problemas conocidos de compatibilidad con GPUs Blackwell (SM120) en modelos DeepSeek y MoE NVFP4.","quote":"Models using multi-head latent attention, such as DeepSeek, may hang or crash during generation on SM120 Blackwell GPUs","source":0},{"text":"Nuevas capacidades de optimización para kernels y gestión de KV cache.","quote":"Add NVFP4 KV cache for DSA and cold-page compression #17681, #18091","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:13:21Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/NVIDIA/TensorRT-LLM/releases/tag/v1.3.0rc26","title":"NVIDIA/TensorRT-LLM v1.3.0rc26","retrieved_at":"2026-10-13T05:51:39Z","sha256":"d0e470d9fd9b8c6f6540bdafdf3a2c2038347e2a78c3b736f8922ea9a25ed9f8","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--nvidia--tensorrt-llm--v1.3.0rc26","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"},"tool":{"slug":"tensorrt-llm","name":"TensorRT-LLM"},"version":"v1.3.0rc26"},{"id":"github:openai/codex:rust-v0.154.0","kind":"software_release","title":"codex 0.154.0","date":"2026-09-09","observed_at":"2026-10-13T06:10:27Z","subject":{"type":"repo","ref":"openai/codex","url":"https://github.com/openai/codex/releases/tag/rust-v0.154.0","entity":"openai","channel":"stable"},"status":"published","severity":"notable","action":"try","tags":[],"affects":["codex"],"summary":{"why":"Esta actualización introduce soporte para el nuevo modelo GPT-6-Astra y añade funcionalidades críticas de gestión de sesiones (worktrees) y mejoras en la infraestructura de ejecución en Windows y macOS."},"highlights":[{"text":"Disponibilidad de GPT-6-Astra en el selector de modelos y catálogos de Amazon Bedrock.","quote":"GPT-6-Astra is now available in the model picker and Amazon Bedrock catalogs.","source":0},{"text":"Soporte experimental para worktrees, permitiendo sesiones aisladas para nuevos checkouts o forks.","quote":"Experimental worktree support lets you create isolated checkouts for new or forked sessions using `--worktree` or `/worktree`","source":0},{"text":"Mejoras en la seguridad del sandbox de macOS contra inyección de entrada en la terminal.","quote":"the macOS sandbox blocks terminal input injection.","source":0},{"text":"Soporte para un servidor Codex de fondo en Windows con gestión de ciclo de vida de demonio.","quote":"Windows sessions can now share a background Codex server, with daemon lifecycle commands and managed updates.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:10:27Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (deprecation)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/openai/codex/releases/tag/rust-v0.154.0","title":"openai/codex 0.154.0","retrieved_at":"2026-10-13T05:51:35Z","sha256":"c78295fc9bda9bcc264e903b1edec3f7181302432a50d7baea4b91c5b7b7ca04","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--openai--codex--rust-v0.154.0","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"},"tool":{"slug":"codex","name":"Codex CLI"},"version":"rust-v0.154.0"},{"id":"github:huggingface/transformers:v5.17.0","kind":"software_release","title":"transformers Release 5.17.0","date":"2026-09-09","observed_at":"2026-10-13T05:56:46Z","subject":{"type":"repo","ref":"huggingface/transformers","url":"https://github.com/huggingface/transformers/releases/tag/v5.17.0","entity":"huggingface","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["breaking","new-model"],"affects":["transformers"],"summary":{"why":"Esta versión introduce soporte para modelos de arquitectura avanzada (HYV4, KimiLinear) y añade múltiples modelos de audio y visión, además de un cambio importante en la implementación de RoPE para modelos de visión que requiere migración."},"highlights":[{"text":"Soporte para el modelo MoE HYV4 de 780B parámetros con ventana de contexto de 1M.","quote":"Hy4-Preview is a 780B-parameter mixture-of-experts language model that activates 49B parameters per token.","source":0},{"text":"Implementación de la arquitectura híbrida KimiLinear con Kimi Delta Attention.","quote":"Kimi Linear is a hybrid linear attention architecture from Moonshot AI","source":0},{"text":"Cambio incompatible en la computación de RoPE para modelos de visión (2D/3D).","quote":"Vision rotary embeddings (2D/3D) have been standardized into a unified RoPE frequency computation module, so users with custom vision models relying on attention-layer-level or model-specific RoPE grid interleaving logic must migrate","source":0},{"text":"Optimización de rendimiento en el proceso de generación al evitar sincronizaciones innecesarias.","quote":"Generation improvements include a performance optimization that avoids unnecessary accelerator synchronization on every decode step","source":0},{"text":"Nuevos modelos de audio: VibeVoice (síntesis de voz), Fun-ASR-Nano (ASR) y Canary (ASR/Traducción).","quote":"Implement VibeVoice","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:56:46Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, deprecation, new-architecture)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/huggingface/transformers/releases/tag/v5.17.0","title":"huggingface/transformers Release 5.17.0","retrieved_at":"2026-10-13T05:51:18Z","sha256":"3a2e3b98ff6a250c5309df8bd8c2e89775ae0f9e0bc1360fb49212cac63fc97d","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--huggingface--transformers--v5.17.0","related":[],"org":{"slug":"huggingface","name":"Hugging Face","kind":"empresa","url":"https://huggingface.co"},"tool":{"slug":"transformers","name":"Transformers"},"version":"v5.17.0"},{"id":"github:vllm-project/vllm:v0.29.0","kind":"software_release","title":"vllm v0.29.0","date":"2026-09-09","observed_at":"2026-10-13T05:56:29Z","subject":{"type":"repo","ref":"vllm-project/vllm","url":"https://github.com/vllm-project/vllm/releases/tag/v0.29.0","entity":"vllm","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["breaking","new-model","performance"],"affects":["vllm"],"summary":{"why":"Es una actualización mayor que cambia el motor de ejecución por defecto (Model Runner V2), introduce soporte para modelos de última generación (DeepSeek V4, Kimi K3, Qwen3) y optimizaciones críticas de rendimiento en arquitecturas Blackwell y Hopper."},"highlights":[{"text":"Model Runner V2 pasa a ser el motor de ejecución por defecto para todos los modelos.","quote":"Model Runner V2 is now the default for all models","source":0},{"text":"Optimización de rendimiento para Blackwell con reducción de latencia del 33.6%.","quote":"Blackwell autotuning with 33.6% E2E latency reduction","source":0}],"facts":[],"verification":{"status":"partial","checked_at":"2026-10-13T05:56:29Z","rejected":["New models: Hy4-preview, Tencent's 770B/49B-active MoE with Gated DeepSeek Sparse Attention and native MTP; Qwen3.8-Flash-Next with BF16/FP8/NVFP4 and MTP; GraniteSWA and GraniteMoeSWA; NemotronH_Omni_Reasoning_V3 with MTP; Kimi K3 NVFP4 checkpoints.","Breaking changes: ten deprecated model architectures removed; python -m vllm.entrypoints.openai.api_server deprecated in favor of vllm serve"]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, deprecation, performance)","llm: publish; 2 citas verificadas, 2 rechazadas"]},"sources":[{"url":"https://github.com/vllm-project/vllm/releases/tag/v0.29.0","title":"vllm-project/vllm v0.29.0","retrieved_at":"2026-10-13T05:51:17Z","sha256":"cbe0e1ba4a357effd1afe1da94a16467bfaa9104e3e805d56e2e68931069b2f7","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--vllm-project--vllm--v0.29.0","related":[],"org":{"slug":"vllm","name":"vLLM","kind":"proyecto","url":"https://vllm.ai"},"tool":{"slug":"vllm","name":"vLLM"},"version":"v0.29.0"},{"id":"rss:cohere-blog:9b6082ebc541","kind":"announcement","title":"Cohere's North Mini Code Megakernel Serving Engine | Cohere","date":"2026-09-08","observed_at":"2026-10-13T06:24:07Z","subject":{"type":"feed","ref":"cohere-blog","url":"https://cohere.com/blog/megakernels","entity":"cohere","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-architecture","performance"],"affects":["vllm"],"summary":{"why":"Cohere presenta un nuevo motor de inferencia optimizado para código que supera significativamente el rendimiento de vLLM mediante el uso de un 'decode megakernel'."},"highlights":[{"text":"Nuevo sistema de servicio de LLM listo para producción basado en un megakernel de decodificación.","quote":"Cohere presents the first production-ready LLM serving system built around a decode megakernel","source":0},{"text":"Mejora de rendimiento de 1.58x respecto a vLLM.","quote":"achieving 1.58× speedup over vLLM","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:24:07Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 2 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://cohere.com/blog/megakernels","title":"Cohere's North Mini Code Megakernel Serving Engine | Cohere","retrieved_at":"2026-10-13T05:52:29Z","sha256":"f986ef81f92cd28ab16943ba8de5af1e1df3f5419db5e3a05694143752f76554","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--cohere-blog--9b6082ebc541","related":[],"org":{"slug":"cohere","name":"Cohere","kind":"laboratorio","url":"https://cohere.com"}},{"id":"hf:nvidia/NV-Reason-CT","kind":"model_release","title":"nvidia publica NV-Reason-CT","date":"2026-09-08","observed_at":"2026-10-13T06:14:49Z","subject":{"type":"model","ref":"nvidia/NV-Reason-CT","url":"https://huggingface.co/nvidia/NV-Reason-CT","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 5.3 B parámetros, licencia openmdw-1.1, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":5322014624,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"openmdw-1.1","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"image-text-to-text","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-08T05:06:51.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"VLM3D_ForConditionalGeneration","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T06:14:49Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/nvidia/NV-Reason-CT","title":"Hugging Face · nvidia/NV-Reason-CT","retrieved_at":"2026-10-13T05:51:50Z","sha256":"9295214ca1cf80171e5a01c03fe87cf3ab80a28b7dad90b626c199eb37ec47b0","license":null},{"url":"https://huggingface.co/nvidia/NV-Reason-CT/blob/main/config.json","title":"nvidia/NV-Reason-CT · config.json","retrieved_at":"2026-10-13T05:51:50Z","sha256":"3f0eb68107bec99000822ab1e58a765548cc1464da9e2c6cb51298737cb9fc61","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"nvidia/NV-Reason-CT","params_total":5322014624,"architecture":"VLM3D_ForConditionalGeneration","context_length":262144,"license":"openmdw-1.1","gated":false,"pipeline":"image-text-to-text","variants":[],"kv_bytes_per_token":32768,"quants":[]},"slug":"hf--nvidia--nv-reason-ct","related":["rss:nvidia-dev:9c631b921733"],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}},{"id":"hf:inclusionAI/LLaDA-UI","kind":"model_release","title":"inclusionAI publica LLaDA-UI","date":"2026-09-06","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"inclusionAI/LLaDA-UI","url":"https://huggingface.co/inclusionAI/LLaDA-UI","entity":"inclusionai","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 16.9 B parámetros, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":16924344064,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"image-text-to-text","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-06T05:08:15.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"LLaDA2MoE_VLForConditionalGeneration","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":8192,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/inclusionAI/LLaDA-UI","title":"Hugging Face · inclusionAI/LLaDA-UI","retrieved_at":"2026-10-13T08:55:33Z","sha256":"331dd49d112d0f59bc08dc054aaa021f03eddbdb1230120ede25b988178c4f06","license":null},{"url":"https://huggingface.co/inclusionAI/LLaDA-UI/blob/main/config.json","title":"inclusionAI/LLaDA-UI · config.json","retrieved_at":"2026-10-13T08:55:33Z","sha256":"14b202fc4506eec43d9c5ea2487ba1f0a9d972143bd08259c47ce0cccd00c54b","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"inclusionAI/LLaDA-UI","params_total":16924344064,"architecture":"LLaDA2MoE_VLForConditionalGeneration","context_length":8192,"license":null,"gated":false,"pipeline":"image-text-to-text","variants":[],"kv_bytes_per_token":40960,"quants":[]},"slug":"hf--inclusionai--llada-ui","related":[],"org":{"slug":"inclusionai","name":"inclusionAI (Ant Group)","kind":"laboratorio","url":"https://github.com/inclusionAI"}},{"id":"hf:inclusionAI/LLaDA2.2-mini","kind":"model_release","title":"inclusionAI publica LLaDA2.2-mini","date":"2026-09-05","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"inclusionAI/LLaDA2.2-mini","url":"https://huggingface.co/inclusionAI/LLaDA2.2-mini","entity":"inclusionai","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 16.3 B parámetros, licencia apache-2.0, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":16255643392,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"apache-2.0","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"text-generation","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-05T05:16:19.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"LLaDA2MoeModelLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":131072,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/inclusionAI/LLaDA2.2-mini","title":"Hugging Face · inclusionAI/LLaDA2.2-mini","retrieved_at":"2026-10-13T08:55:33Z","sha256":"23619debc1fa6d03702e310b7620d588be01fd61bdd2db6e68a67dc3eb38a659","license":null},{"url":"https://huggingface.co/inclusionAI/LLaDA2.2-mini/blob/main/config.json","title":"inclusionAI/LLaDA2.2-mini · config.json","retrieved_at":"2026-10-13T08:55:33Z","sha256":"34bef4d06e42f6c6c45b5dae9f59a9c36dc511c2ff66f7486283cdc384c593d2","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"inclusionAI/LLaDA2.2-mini","params_total":16255643392,"architecture":"LLaDA2MoeModelLM","context_length":131072,"license":"apache-2.0","gated":false,"pipeline":"text-generation","variants":[],"kv_bytes_per_token":40960,"quants":[]},"slug":"hf--inclusionai--llada2.2-mini","related":[],"org":{"slug":"inclusionai","name":"inclusionAI (Ant Group)","kind":"laboratorio","url":"https://github.com/inclusionAI"}},{"id":"github:sgl-project/sglang:v0.5.19","kind":"software_release","title":"sglang v0.5.19","date":"2026-09-05","observed_at":"2026-10-13T06:12:53Z","subject":{"type":"repo","ref":"sgl-project/sglang","url":"https://github.com/sgl-project/sglang/releases/tag/v0.5.19","entity":"sglang","channel":"stable"},"status":"published","severity":"notable","action":"try","tags":["performance","security"],"affects":["sglang"],"summary":{"why":"Esta versión de SGLang introduce soporte para nuevos modelos (Qwen3.8, Granite 4.2), optimizaciones críticas de rendimiento en hardware Blackwell y Hopper, y la implementación de beam search."},"highlights":[{"text":"Soporte para nuevos modelos autoregresivos y de difusión","quote":"New models in this release (see the [cookbook](https://docs.sglang.io/cookbook) for all supported models):","source":0},{"text":"Implementación de beam search para obtener múltiples secuencias","quote":"SGLang can now do beam search. Pass `beam_width` in your request and you get back the `n` best sequences instead of a single sample.","source":0},{"text":"Optimización de rendimiento para arquitecturas Hopper (W4A8 MoE)","quote":"DeepSeek-V4-Flash gains about 12% output throughput with no change in GSM8K accuracy.","source":0},{"text":"Mejoras de rendimiento en kernels de prefill para Blackwell","quote":"DSA prefill top-k moves to the v2 kernel, 1.3 to 1.8 times faster on B200","source":0},{"text":"Optimización de atención para hardware AMD (MI300X/MI355X)","quote":"The new persistent Lean kernel spreads the work across all of them, for up to 1.52x more throughput","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:12:53Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, deprecation, feature, performance)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/sgl-project/sglang/releases/tag/v0.5.19","title":"sgl-project/sglang v0.5.19","retrieved_at":"2026-10-13T05:51:38Z","sha256":"58f6fe68f3e3d8d4317704b322de45a884f3c2782d97fc4233b693e5bb44c736","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--sgl-project--sglang--v0.5.19","related":[],"org":{"slug":"sglang","name":"SGLang","kind":"proyecto","url":"https://sglang.ai"},"tool":{"slug":"sglang","name":"SGLang"},"version":"v0.5.19"},{"id":"hf:inclusionAI/Ling-3.0-flash-VL","kind":"model_release","title":"inclusionAI publica Ling-3.0-flash-VL","date":"2026-09-04","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"inclusionAI/Ling-3.0-flash-VL","url":"https://huggingface.co/inclusionAI/Ling-3.0-flash-VL","entity":"inclusionai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Modelo nuevo: 124.8 B parámetros, licencia mit, 26 cuantizaciones GGUF disponibles."},"highlights":[],"facts":[{"field":"params_total","value":124848460496,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"mit","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"image-text-to-text","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-04T15:24:37.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"BailingMoeV3VLForConditionalGeneration","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":131072,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}},{"field":"gguf_quants","value":26,"evidence":{"source":2,"locator":"siblings[].size"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/inclusionAI/Ling-3.0-flash-VL","title":"Hugging Face · inclusionAI/Ling-3.0-flash-VL","retrieved_at":"2026-10-13T08:55:34Z","sha256":"689a7a42606e19753f7fa8762bcef75f7cd1e4bc68ac2110414c0693eef9a387","license":null},{"url":"https://huggingface.co/inclusionAI/Ling-3.0-flash-VL/blob/main/config.json","title":"inclusionAI/Ling-3.0-flash-VL · config.json","retrieved_at":"2026-10-13T08:55:34Z","sha256":"34e3989b488f45da0cce2a69cf4cc7907e85487899306623c80737a48b88aeff","license":null},{"url":"https://huggingface.co/bartowski/Ling-3.0-flash-VL-GGUF","title":"Hugging Face · bartowski/Ling-3.0-flash-VL-GGUF","retrieved_at":"2026-10-13T08:55:34Z","sha256":"5fe0b39a7440afd37d3874073f82618ed3707da34cc9aa54885edab16fc18d81","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"inclusionAI/Ling-3.0-flash-VL","params_total":124848460496,"architecture":"BailingMoeV3VLForConditionalGeneration","context_length":131072,"license":"mit","gated":false,"pipeline":"image-text-to-text","variants":[],"kv_bytes_per_token":48384,"quants":[{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"IQ1_S","bytes":28065095392},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"IQ1_M","bytes":31300674272},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"IQ2_XXS","bytes":36692503520},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"IQ2_XS","bytes":38604598240},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"IQ2_S","bytes":41006820320},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"IQ2_M","bytes":45320276960},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"Q2_K","bytes":47805205472},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"IQ3_XXS","bytes":53936622752},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"IQ3_XS","bytes":56830692512},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"Q3_K_S","bytes":56833445024},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"Q3_K_M","bytes":59766836384},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"Q3_K_L","bytes":63345102016},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"IQ3_M","bytes":67612675232},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"IQ4_XS","bytes":69374282912},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"Q4_0","bytes":71111168160},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"Q4_K_S","bytes":73977662656},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"Q4_1","bytes":78473796768},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"Q4_K_M","bytes":78661520544},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"IQ4_NL","bytes":78677773472},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"Q4_K_L","bytes":84882197792},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"Q5_K_S","bytes":88955915520},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"Q5_K_M","bytes":95884381472},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"Q6_K_S","bytes":104574455072},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"Q6_K","bytes":109694127360},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"Q8_0","bytes":132365277248},{"repo":"bartowski/Ling-3.0-flash-VL-GGUF","quant":"BF16","bytes":248944047584}]},"slug":"hf--inclusionai--ling-3.0-flash-vl","related":[],"org":{"slug":"inclusionai","name":"inclusionAI (Ant Group)","kind":"laboratorio","url":"https://github.com/inclusionAI"}},{"id":"rss:devblogs-sk:504d112a9f08","kind":"announcement","title":"Native memory for Microsoft Agent Framework with Azure Cosmos DB","date":"2026-09-04","observed_at":"2026-10-13T06:22:39Z","subject":{"type":"feed","ref":"devblogs-sk","url":"https://devblogs.microsoft.com/agent-framework/native-memory-for-microsoft-agent-framework-with-azure-cosmos-db/","entity":"microsoft","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-architecture","new-model"],"affects":[],"summary":{"why":"Permite a los agentes de Microsoft tener memoria persistente y duradera entre sesiones utilizando Azure Cosmos DB, facilitando la creación de agentes con contexto histórico."},"highlights":[{"text":"Nueva integración en preview para memoria duradera y entre sesiones mediante Azure Cosmos DB.","quote":"Today, we’re announcing a new preview integration that gives Microsoft Agent Framework agents durable, cross-session memory backed by Azure Cosmos DB.","source":0},{"text":"Lanzamiento del nuevo paquete de Python para gestionar este contexto de memoria.","quote":"The new Python package, agent-framework-azure-cosmos-memory, provides CosmosMemoryContextProvider.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:22:39Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 2 citas verificadas, 0 rechazadas","publish con severidad minor: queda en observación","corrección editorial: lanzamiento real nunca es minor (la severidad minor era un fallo de calibración del prompt)"]},"sources":[{"url":"https://devblogs.microsoft.com/agent-framework/native-memory-for-microsoft-agent-framework-with-azure-cosmos-db/","title":"Native memory for Microsoft Agent Framework with Azure Cosmos DB","retrieved_at":"2026-10-13T05:51:56Z","sha256":"c496d272eca799583c83b404bb557b20561e86072cbc29b460fd8a12121ba0a4","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--devblogs-sk--504d112a9f08","related":[],"org":{"slug":"microsoft","name":"Microsoft","kind":"empresa","url":"https://www.microsoft.com"}},{"id":"hf:inclusionAI/Ling-3.0-flash-Fin","kind":"model_release","title":"inclusionAI publica Ling-3.0-flash-Fin","date":"2026-09-03","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"inclusionAI/Ling-3.0-flash-Fin","url":"https://huggingface.co/inclusionAI/Ling-3.0-flash-Fin","entity":"inclusionai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Modelo nuevo: 127.5 B parámetros, licencia mit, 28 cuantizaciones GGUF disponibles."},"highlights":[],"facts":[{"field":"params_total","value":127486405600,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"mit","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"text-generation","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-03T06:09:53.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"BailingMoeV3ForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}},{"field":"gguf_quants","value":28,"evidence":{"source":2,"locator":"siblings[].size"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/inclusionAI/Ling-3.0-flash-Fin","title":"Hugging Face · inclusionAI/Ling-3.0-flash-Fin","retrieved_at":"2026-10-13T08:55:34Z","sha256":"dffed1fa2000a9b8d02b18fa3ccfb0e0e171858724386f61e7ef7c10323a3359","license":null},{"url":"https://huggingface.co/inclusionAI/Ling-3.0-flash-Fin/blob/main/config.json","title":"inclusionAI/Ling-3.0-flash-Fin · config.json","retrieved_at":"2026-10-13T08:55:35Z","sha256":"04a3acd13eaedccd7115dbd3717837e84d70e2d0954888c184c8d073f82645c5","license":null},{"url":"https://huggingface.co/bartowski/Ling-3.0-flash-Fin-GGUF","title":"Hugging Face · bartowski/Ling-3.0-flash-Fin-GGUF","retrieved_at":"2026-10-13T08:55:35Z","sha256":"f73ca8a828d5ab3a84d1b7cba3c6abbca68772f556cd7cdfc75b86a0dc383fd2","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"inclusionAI/Ling-3.0-flash-Fin","params_total":127486405600,"architecture":"BailingMoeV3ForCausalLM","context_length":262144,"license":"mit","gated":false,"pipeline":"text-generation","variants":[],"kv_bytes_per_token":48384,"quants":[{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"IQ1_S","bytes":27571028224},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"IQ1_M","bytes":30561301888},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"IQ2_XXS","bytes":35238382848},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"IQ2_XS","bytes":39036670720},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"IQ2_S","bytes":39698819840},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"IQ2_M","bytes":43638665472},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q2_K","bytes":45900759680},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q2_K_L","bytes":46293719680},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"IQ3_XXS","bytes":54118505952},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q3_K_S","bytes":56287608544},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"IQ3_XS","bytes":58941685440},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q3_K_M","bytes":58969587424},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q3_K_L","bytes":61427678944},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"IQ3_M","bytes":61517496032},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q3_K_XL","bytes":61779771104},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"IQ4_XS","bytes":68739201632},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"IQ4_NL","bytes":72564891744},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q4_0","bytes":72841125952},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q4_K_S","bytes":75049689184},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q4_K_M","bytes":77804990560},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q4_K_L","bytes":78103640160},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q4_1","bytes":80204410080},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q5_K_S","bytes":87892609216},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q5_K_M","bytes":90661271776},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q5_K_L","bytes":90909622496},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q6_K","bytes":109259389504},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"Q8_0","bytes":135633373216},{"repo":"bartowski/Ling-3.0-flash-Fin-GGUF","quant":"BF16","bytes":255091083616}]},"slug":"hf--inclusionai--ling-3.0-flash-fin","related":[],"org":{"slug":"inclusionai","name":"inclusionAI (Ant Group)","kind":"laboratorio","url":"https://github.com/inclusionAI"}},{"id":"rss:openai:039a4975e102","kind":"announcement","title":"GPT-6 Astra: A new generation of intelligence","date":"2026-09-03","observed_at":"2026-10-13T06:17:32Z","subject":{"type":"feed","ref":"openai","url":"https://openai.com/index/gpt-6-astra","entity":"openai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":[],"affects":[],"summary":{"why":"Anuncio de una nueva generación de modelos con capacidades avanzadas en codificación y uso de computadoras, lo que implica un nuevo estándar de inteligencia para integrar en flujos de trabajo."},"highlights":[{"text":"Introducción de GPT-6 Astra con capacidades de vanguardia en uso de computadoras, programación, ciberseguridad y ciencia.","quote":"state-of-the-art capabilities across computer use, coding, cybersecurity, and science","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:17:32Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://openai.com/index/gpt-6-astra","title":"GPT-6 Astra: A new generation of intelligence","retrieved_at":"2026-10-13T05:51:53Z","sha256":"298d4495d9399e11b6721659d0931db9cba451be14c2d7132fa578ed216dd62e","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--openai--039a4975e102","related":[],"org":{"slug":"openai","name":"OpenAI","kind":"laboratorio","url":"https://openai.com"}},{"id":"github:anthropics/claude-code:v2.1.260","kind":"software_release","title":"claude-code v2.1.260","date":"2026-09-03","observed_at":"2026-10-13T05:59:50Z","subject":{"type":"repo","ref":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.260","entity":"anthropic","channel":"stable"},"status":"published","severity":"notable","action":"update","tags":["performance","security"],"affects":["claude-code"],"summary":{"why":"Esta actualización corrige vulnerabilidades críticas en las reglas de permisos del sandbox de Bash (permitiendo escrituras en carpetas de solo lectura) y mejora significativamente el manejo de caché de prompts y el uso de CPU en sesiones SDK."},"highlights":[{"text":"Corrección de seguridad en las reglas de permisos de Bash que permitían escribir en carpetas de solo lectura.","quote":"Fixed `Edit`/`Write`/`Read` permission rules whose path contains parentheses being dropped as invalid or ignored by the Bash sandbox, which left \"read-only\" folders writable","source":0},{"text":"Mejoras en la gestión de caché de prompts para evitar reenvíos innecesarios de contexto.","quote":"Fixed prompt caching on Claude Fable 5.1 not covering the context attached after tool results, so it was re-sent as uncached input on every tool-call turn","source":0},{"text":"Optimización del uso de CPU para sesiones no interactivas (SDK/headless).","quote":"Improved idle CPU usage of non-interactive (`-p` / SDK) sessions","source":0},{"text":"Nueva funcionalidad de panel de diff para visualizar cambios no confirmados durante la edición.","quote":"Added a diff panel that opens beside the conversation in fullscreen mode and shows your uncommitted changes as Claude edits; toggle it with `/diff`","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T05:59:50Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (feature)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.260","title":"anthropics/claude-code v2.1.260","retrieved_at":"2026-10-13T05:51:19Z","sha256":"74b6611e274cbe30aad0f14051cbfc108ad70671f58cfe3f702f5345fc7069fe","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--anthropics--claude-code--v2.1.260","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"},"tool":{"slug":"claude-code","name":"Claude Code"},"version":"v2.1.260"},{"id":"hf:microsoft/VibeVoice-ASR-Streaming-1.5B","kind":"model_release","title":"microsoft publica VibeVoice-ASR-Streaming-1.5B","date":"2026-09-02","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"microsoft/VibeVoice-ASR-Streaming-1.5B","url":"https://huggingface.co/microsoft/VibeVoice-ASR-Streaming-1.5B","entity":"microsoft","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 2.8 B parámetros, licencia mit, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":2814116321,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"mit","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"automatic-speech-recognition","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-02T15:51:38.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"VibeVoiceForASRStreamingTraining","evidence":{"source":1,"locator":"config.json#architectures[0]"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/microsoft/VibeVoice-ASR-Streaming-1.5B","title":"Hugging Face · microsoft/VibeVoice-ASR-Streaming-1.5B","retrieved_at":"2026-10-13T08:54:56Z","sha256":"db85a0ee7c8a8d995540bb319ca8c98f1a5f19a7f931ef66dddb8f4ea2fbd612","license":null},{"url":"https://huggingface.co/microsoft/VibeVoice-ASR-Streaming-1.5B/blob/main/config.json","title":"microsoft/VibeVoice-ASR-Streaming-1.5B · config.json","retrieved_at":"2026-10-13T08:54:56Z","sha256":"19c694822174941dbf93fe62cb887596769b5c03df24f7db8346e1bc883e5ae6","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"microsoft/VibeVoice-ASR-Streaming-1.5B","params_total":2814116321,"architecture":"VibeVoiceForASRStreamingTraining","context_length":null,"license":"mit","gated":false,"pipeline":"automatic-speech-recognition","variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--microsoft--vibevoice-asr-streaming-1.5b","related":[],"org":{"slug":"microsoft","name":"Microsoft","kind":"empresa","url":"https://www.microsoft.com"}},{"id":"hf:microsoft/VibeVoice-ASR-Streaming-7B","kind":"model_release","title":"microsoft publica VibeVoice-ASR-Streaming-7B","date":"2026-09-02","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"microsoft/VibeVoice-ASR-Streaming-7B","url":"https://huggingface.co/microsoft/VibeVoice-ASR-Streaming-7B","entity":"microsoft","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 8.7 B parámetros, licencia mit, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":8674021857,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"mit","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"automatic-speech-recognition","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-02T15:46:22.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"VibeVoiceForASRStreamingTraining","evidence":{"source":1,"locator":"config.json#architectures[0]"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/microsoft/VibeVoice-ASR-Streaming-7B","title":"Hugging Face · microsoft/VibeVoice-ASR-Streaming-7B","retrieved_at":"2026-10-13T08:54:57Z","sha256":"107b4eb727e3a16855c527fe9453aa5e16600e4ec52698353521720dac84f57d","license":null},{"url":"https://huggingface.co/microsoft/VibeVoice-ASR-Streaming-7B/blob/main/config.json","title":"microsoft/VibeVoice-ASR-Streaming-7B · config.json","retrieved_at":"2026-10-13T08:54:57Z","sha256":"546cd36ffda33330e9838f8329d987293e7bf819435329419dc0a85245627aa3","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"microsoft/VibeVoice-ASR-Streaming-7B","params_total":8674021857,"architecture":"VibeVoiceForASRStreamingTraining","context_length":null,"license":"mit","gated":false,"pipeline":"automatic-speech-recognition","variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--microsoft--vibevoice-asr-streaming-7b","related":[],"org":{"slug":"microsoft","name":"Microsoft","kind":"empresa","url":"https://www.microsoft.com"}},{"id":"hf:nvidia/Nemotron-3-Labs-Ultra-Math-RL","kind":"model_release","title":"nvidia publica Nemotron-3-Labs-Ultra-Math-RL","date":"2026-09-02","observed_at":"2026-10-13T06:14:49Z","subject":{"type":"model","ref":"nvidia/Nemotron-3-Labs-Ultra-Math-RL","url":"https://huggingface.co/nvidia/Nemotron-3-Labs-Ultra-Math-RL","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 560.5 B parámetros, licencia other:openmdw-1.1, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":560524578816,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"other:openmdw-1.1","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"text-generation","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-02T19:11:33.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"NemotronHForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T06:14:49Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/nvidia/Nemotron-3-Labs-Ultra-Math-RL","title":"Hugging Face · nvidia/Nemotron-3-Labs-Ultra-Math-RL","retrieved_at":"2026-10-13T05:51:51Z","sha256":"401f7d7a42ae9b7c0b5ec68c43ebfb256f869b72204e86993303b8e7fcdc2426","license":null},{"url":"https://huggingface.co/nvidia/Nemotron-3-Labs-Ultra-Math-RL/blob/main/config.json","title":"nvidia/Nemotron-3-Labs-Ultra-Math-RL · config.json","retrieved_at":"2026-10-13T05:51:51Z","sha256":"98d967050e0df1059e3bbde2576738ad49325498c9d27f4fbb109bc75196ad4b","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"nvidia/Nemotron-3-Labs-Ultra-Math-RL","params_total":560524578816,"architecture":"NemotronHForCausalLM","context_length":262144,"license":"other:openmdw-1.1","gated":false,"pipeline":"text-generation","variants":[],"kv_bytes_per_token":110592,"quants":[]},"slug":"hf--nvidia--nemotron-3-labs-ultra-math-rl","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}},{"id":"hf:nvidia/Nemotron-3-Labs-Ultra-Math-SFT","kind":"model_release","title":"nvidia publica Nemotron-3-Labs-Ultra-Math-SFT","date":"2026-09-02","observed_at":"2026-10-13T06:14:49Z","subject":{"type":"model","ref":"nvidia/Nemotron-3-Labs-Ultra-Math-SFT","url":"https://huggingface.co/nvidia/Nemotron-3-Labs-Ultra-Math-SFT","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 560.5 B parámetros, licencia other:openmdw-1.1, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":560524603904,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"other:openmdw-1.1","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"text-generation","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-02T19:11:32.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"NemotronHForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T06:14:49Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/nvidia/Nemotron-3-Labs-Ultra-Math-SFT","title":"Hugging Face · nvidia/Nemotron-3-Labs-Ultra-Math-SFT","retrieved_at":"2026-10-13T05:51:51Z","sha256":"1c90a97345d8fb2f7bdd6ed83291ac43372d051303fa560c5a228308ea78cc5c","license":null},{"url":"https://huggingface.co/nvidia/Nemotron-3-Labs-Ultra-Math-SFT/blob/main/config.json","title":"nvidia/Nemotron-3-Labs-Ultra-Math-SFT · config.json","retrieved_at":"2026-10-13T05:51:51Z","sha256":"b84072d71664b3c2bec5233a2192d80dba8e58a0ced1360853254e4f10511947","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"nvidia/Nemotron-3-Labs-Ultra-Math-SFT","params_total":560524603904,"architecture":"NemotronHForCausalLM","context_length":262144,"license":"other:openmdw-1.1","gated":false,"pipeline":"text-generation","variants":[],"kv_bytes_per_token":110592,"quants":[]},"slug":"hf--nvidia--nemotron-3-labs-ultra-math-sft","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}},{"id":"rss:hf-blog:91dad0068ec2","kind":"announcement","title":"Introducing @huggingface/kernels: 200+ WebGPU Kernels for Local AI","date":"2026-09-01","observed_at":"2026-10-13T06:18:39Z","subject":{"type":"feed","ref":"hf-blog","url":"https://huggingface.co/blog/webgpu-kernels","entity":"huggingface","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-architecture","performance"],"affects":[],"summary":{"why":"Hugging Face lanza una colección de kernels optimizados para WebGPU, permitiendo ejecutar modelos de IA directamente en el navegador con alto rendimiento."},"highlights":[{"text":"Más de 200 kernels de WebGPU para IA local.","quote":"200+ WebGPU Kernels for Local AI","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:18:39Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://huggingface.co/blog/webgpu-kernels","title":"Introducing @huggingface/kernels: 200+ WebGPU Kernels for Local AI","retrieved_at":"2026-10-13T05:51:54Z","sha256":"0d68325b4ac628ba11171912d761f26c9c7ff4b35b1eeff36252ffaaf941efc6","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--hf-blog--91dad0068ec2","related":[],"org":{"slug":"huggingface","name":"Hugging Face","kind":"empresa","url":"https://huggingface.co"}},{"id":"hf:nvidia/NVIDIA-NemotronLabs-AI-for-Media-Sports-Tennis","kind":"model_release","title":"nvidia publica NVIDIA-NemotronLabs-AI-for-Media-Sports-Tennis","date":"2026-09-01","observed_at":"2026-10-13T06:14:49Z","subject":{"type":"model","ref":"nvidia/NVIDIA-NemotronLabs-AI-for-Media-Sports-Tennis","url":"https://huggingface.co/nvidia/NVIDIA-NemotronLabs-AI-for-Media-Sports-Tennis","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 33.0 B parámetros, licencia other:openmdw-1.1, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":33015598912,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"other:openmdw-1.1","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"any-to-any","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-09-01T19:58:31.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"NemotronH_Nano_Omni_Reasoning_V3","evidence":{"source":1,"locator":"config.json#architectures[0]"}}],"verification":{"status":"verified","checked_at":"2026-10-13T06:14:49Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/nvidia/NVIDIA-NemotronLabs-AI-for-Media-Sports-Tennis","title":"Hugging Face · nvidia/NVIDIA-NemotronLabs-AI-for-Media-Sports-Tennis","retrieved_at":"2026-10-13T05:51:52Z","sha256":"32643880b9875eb0d423b5f548ecd6cd1473c91e192fbdee12a77ddf86b07720","license":null},{"url":"https://huggingface.co/nvidia/NVIDIA-NemotronLabs-AI-for-Media-Sports-Tennis/blob/main/config.json","title":"nvidia/NVIDIA-NemotronLabs-AI-for-Media-Sports-Tennis · config.json","retrieved_at":"2026-10-13T05:51:52Z","sha256":"282e180fb1fa2e501ec57c7032b7f39918c36f2a8661b3b2d717af6379a8a98d","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"nvidia/NVIDIA-NemotronLabs-AI-for-Media-Sports-Tennis","params_total":33015598912,"architecture":"NemotronH_Nano_Omni_Reasoning_V3","context_length":null,"license":"other:openmdw-1.1","gated":false,"pipeline":"any-to-any","variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--nvidia--nvidia-nemotronlabs-ai-for-media-sports-tennis","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}},{"id":"github:anthropics/claude-code:v2.1.257","kind":"software_release","title":"claude-code v2.1.257","date":"2026-09-01","observed_at":"2026-10-13T06:00:04Z","subject":{"type":"repo","ref":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.257","entity":"anthropic","channel":"stable"},"status":"published","severity":"notable","action":"try","tags":["new-model","performance","security"],"affects":["claude-code"],"summary":{"why":"Esta versión introduce cambios de seguridad relevantes (nuevas reglas de contención para metadata/egress, restricción de lectura de archivos fuera del directorio de trabajo, cierre de fugas de credenciales en gateways) y un nuevo modelo por defecto (Fable 5.1). También incluye mejoras de rendimiento y correcciones de bugs críticos en sesiones en segundo plano y sandboxing. Merece la pena actualiza"},"highlights":[{"text":"Nuevo modelo por defecto: Claude Fable 5.1 con contexto de 1M tokens.","quote":"Added Claude Fable 5.1 (`claude-fable-5-1`), now the default Fable model — 1M context","source":0},{"text":"Nueva regla de seguridad en modo auto para bloquear intentos de escape de contención (metadata, evasión de egress).","quote":"Added a Containment Escape rule to auto mode so cloud metadata-credential fetches, egress evasion, and cross-tenant reach are no longer auto-approved","source":0},{"text":"Bloqueo de lectura de archivos fuera del directorio de trabajo mediante prompt único y configuración.","quote":"Added a one-time prompt in auto mode before the first file read outside the working directories, with the option to block such reads","source":0},{"text":"Mejora de rendimiento: menos re-renderizado por turno y streaming más fluido en conversaciones largas.","quote":"Improved rendering performance: less re-render work per turn in long conversations, streaming no longer slows down as the reply grows","source":0},{"text":"Corrección de fugas de credenciales y problemas de autenticación en gateways y proveedores cloud (Bedrock, Vertex).","quote":"Fixed a doubly-listed custom `Authorization` header overriding the configured credential on Bedrock, Mantle, Vertex, and WIF","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:00:04Z","rejected":[]},"triage":{"method":"llm","model":"qwen3.6","reasons":["reglas: publish (feature, new-architecture)","llm: publish; 5 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/anthropics/claude-code/releases/tag/v2.1.257","title":"anthropics/claude-code v2.1.257","retrieved_at":"2026-10-13T05:51:19Z","sha256":"29a9b24dce49cfb7b9910e0559f3e6ac49072945c5f3ff11b95e70f4c071f8b6","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--anthropics--claude-code--v2.1.257","related":[],"org":{"slug":"anthropic","name":"Anthropic","kind":"laboratorio","url":"https://www.anthropic.com"},"tool":{"slug":"claude-code","name":"Claude Code"},"version":"v2.1.257"},{"id":"rss:ollama-blog:8b2f74d36b50","kind":"announcement","title":"Ollama's transparent pricing","date":"2026-08-31","observed_at":"2026-10-13T06:20:48Z","subject":{"type":"feed","ref":"ollama-blog","url":"https://ollama.com/blog/transparent-pricing","entity":"ollama","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["license","pricing"],"affects":["ollama"],"summary":{"why":"Cambio en el modelo de facturación de los planes de Ollama hacia un esquema basado en tokens, lo que afecta la previsibilidad de costes para usuarios de planes Pro, Max y Team."},"highlights":[{"text":"Los planes Pro, Max y Team de Ollama ahora utilizan precios por token estándar de la industria.","quote":"Ollama's Pro, Max, and Team plans now use industry-standard per-token pricing with usage included on every plan.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:20:48Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: hold (sin señales)","llm: publish; 1 citas verificadas, 0 rechazadas","publish con severidad minor: queda en observación","corrección editorial: lanzamiento real nunca es minor (la severidad minor era un fallo de calibración del prompt)"]},"sources":[{"url":"https://ollama.com/blog/transparent-pricing","title":"Ollama's transparent pricing","retrieved_at":"2026-10-13T05:51:54Z","sha256":"657ffb7098f5b1143df1a00c0f15490d86143331eac3eb6e481cca9c23aad645","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"rss--ollama-blog--8b2f74d36b50","related":[],"org":{"slug":"ollama","name":"Ollama","kind":"proyecto","url":"https://ollama.com"}},{"id":"hf:deepseek-ai/DeepSeek-V4-Flash-Vision-Exp","kind":"model_release","title":"deepseek-ai publica DeepSeek-V4-Flash-Vision-Exp","date":"2026-08-31","observed_at":"2026-10-13T06:14:49Z","subject":{"type":"model","ref":"deepseek-ai/DeepSeek-V4-Flash-Vision-Exp","url":"https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp","entity":"deepseek-ai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Modelo nuevo: 304.6 B parámetros, licencia mit, 10 cuantizaciones GGUF disponibles."},"highlights":[],"facts":[{"field":"params_total","value":304646824126,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"mit","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"image-text-to-text","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-31T06:16:18.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"DeepseekV4ForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":1048576,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}},{"field":"gguf_quants","value":10,"evidence":{"source":2,"locator":"siblings[].size"}}],"verification":{"status":"verified","checked_at":"2026-10-13T06:14:49Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp","title":"Hugging Face · deepseek-ai/DeepSeek-V4-Flash-Vision-Exp","retrieved_at":"2026-10-13T05:51:47Z","sha256":"3901ed7bd8bf55621cb6a2318351a3a140878633e44c3eb28653e72851cc4985","license":null},{"url":"https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp/blob/main/config.json","title":"deepseek-ai/DeepSeek-V4-Flash-Vision-Exp · config.json","retrieved_at":"2026-10-13T05:51:48Z","sha256":"a23909345adfb3f10aaa9accc52a030e56b7c932c418dbeec738f388174ada56","license":null},{"url":"https://huggingface.co/unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF","title":"Hugging Face · unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF","retrieved_at":"2026-10-13T05:51:48Z","sha256":"853126acfc576ecd6df02469d560b4452112f9fdbd9203acbc155f83c86f8e26","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"deepseek-ai/DeepSeek-V4-Flash-Vision-Exp","params_total":304646824126,"architecture":"DeepseekV4ForCausalLM","context_length":1048576,"license":"mit","gated":false,"pipeline":"image-text-to-text","variants":[],"kv_bytes_per_token":88064,"quants":[{"repo":"unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF","quant":"IQ1_S","bytes":82438622112},{"repo":"unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF","quant":"IQ1_M","bytes":86867807104},{"repo":"unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF","quant":"IQ2_XXS","bytes":90693012352},{"repo":"unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF","quant":"Q2_K_XL","bytes":96832555936},{"repo":"unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF","quant":"IQ3_XXS","bytes":102999936000},{"repo":"unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF","quant":"IQ3_S","bytes":114391665664},{"repo":"unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF","quant":"Q3_K_XL","bytes":128206777376},{"repo":"unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF","quant":"IQ4_XS","bytes":136662494240},{"repo":"unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF","quant":"Q4_K_XL","bytes":155095288672},{"repo":"unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF","quant":"Q8_K_XL","bytes":161869663072}]},"slug":"hf--deepseek-ai--deepseek-v4-flash-vision-exp","related":[],"org":{"slug":"deepseek","name":"DeepSeek","kind":"laboratorio","url":"https://www.deepseek.com"}},{"id":"github:open-webui/open-webui:v0.11.2","kind":"software_release","title":"open-webui v0.11.2","date":"2026-08-31","observed_at":"2026-10-13T06:14:17Z","subject":{"type":"repo","ref":"open-webui/open-webui","url":"https://github.com/open-webui/open-webui/releases/tag/v0.11.2","entity":"open-webui","channel":"auto-build"},"status":"published","severity":"critical","action":"update","tags":["security"],"affects":["open-webui"],"summary":{"why":"Esta versión incluye correcciones críticas de seguridad y control de acceso, por lo que se recomienda la actualización inmediata en entornos de producción."},"highlights":[{"text":"Correcciones de seguridad y control de acceso.","quote":"This release includes security and access-control fixes. We recommend updating production deployments at your earliest convenience.","source":0},{"text":"Nuevo paso de filtro de solicitudes para ajustar el payload antes de la llamada al modelo.","quote":"Filter authors can now use `request` to adjust the payload right before each model call","source":0},{"text":"Mejora en el streaming para modelos de razonamiento (reasoning models).","quote":"A reply from a reasoning model now streams through to the end instead of showing its first few words and then freezing until generation finishes.","source":0},{"text":"Reducción de sobrecarga en mensajes cuando no hay pipelines configurados.","quote":"Deployments without pipelines configured, which is the default, no longer pay setup work for them on each chat message and background task.","source":0}],"facts":[],"verification":{"status":"verified","checked_at":"2026-10-13T06:14:17Z","rejected":[]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, security)","llm: publish; 4 citas verificadas, 0 rechazadas"]},"sources":[{"url":"https://github.com/open-webui/open-webui/releases/tag/v0.11.2","title":"open-webui/open-webui v0.11.2","retrieved_at":"2026-10-13T05:51:43Z","sha256":"0ee11a46d12a117468b12fc6b2cca08b45d5c5c29e95dd3f7df5cd6dbc621456","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--open-webui--open-webui--v0.11.2","related":[],"org":{"slug":"open-webui","name":"Open WebUI","kind":"proyecto","url":"https://openwebui.com"},"tool":{"slug":"open-webui","name":"Open WebUI"},"version":"v0.11.2"},{"id":"github:NVIDIA/TensorRT-LLM:v1.3.0rc25","kind":"software_release","title":"TensorRT-LLM v1.3.0rc25","date":"2026-08-31","observed_at":"2026-10-13T06:13:26Z","subject":{"type":"repo","ref":"NVIDIA/TensorRT-LLM","url":"https://github.com/NVIDIA/TensorRT-LLM/releases/tag/v1.3.0rc25","entity":"nvidia","channel":"auto-build"},"status":"published","severity":"notable","action":"watch","tags":["breaking","new-model","performance"],"affects":["tensorrt-llm"],"summary":{"why":"Esta versión introduce el KV Cache Manager V2 por defecto para modelos clave (DeepSeek, Gemma, Qwen), lo que mejora la escalabilidad pero conlleva riesgos de estabilidad según la lista de 'Known Issues'. Además, añade soporte para modelos muy recientes como Gemma 4 y DeepSeek V4."},"highlights":[{"text":"Soporte para nuevos modelos y arquitecturas como Cosmos3, HunyuanVideo y Kimi K3.","quote":"Add Cosmos3-Edge support for text-to-image, text-to-video, and image-to-video generation","source":0},{"text":"Nuevas capacidades de API para generación visual y métricas de carga de pesos.","quote":"Add the `/v1/images/edits` endpoint for visual-generation models","source":0},{"text":"Advertencia de inestabilidad: posibles bloqueos en la generación de Gemma 4 y fallos en GPUs B200.","quote":"Generation may hang indefinitely across multiple models, including Gemma 4.","source":0}],"facts":[],"verification":{"status":"partial","checked_at":"2026-10-13T06:13:26Z","rejected":["Enabled KV cache manager V2 by default for: DeepSeek V3, DeepSeek R1, and DeepSeek V3 Lite... Gemma 3 and Gemma 4"]},"triage":{"method":"llm","model":"gemma4","reasons":["reglas: publish (breaking, deprecation)","llm: publish; 3 citas verificadas, 1 rechazadas"]},"sources":[{"url":"https://github.com/NVIDIA/TensorRT-LLM/releases/tag/v1.3.0rc25","title":"NVIDIA/TensorRT-LLM v1.3.0rc25","retrieved_at":"2026-10-13T05:51:39Z","sha256":"02e2525750033e197165dd4c9e5b947e83968f7b152f6134520d433be1f8aeb0","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"slug":"github--nvidia--tensorrt-llm--v1.3.0rc25","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"},"tool":{"slug":"tensorrt-llm","name":"TensorRT-LLM"},"version":"v1.3.0rc25"},{"id":"hf:microsoft/rho-libero","kind":"model_release","title":"microsoft publica rho-libero","date":"2026-08-28","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"microsoft/rho-libero","url":"https://huggingface.co/microsoft/rho-libero","entity":"microsoft","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 5.2 B parámetros, licencia mit, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":5247548896,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"mit","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"created_at","value":"2026-08-28T23:17:11.000Z","evidence":{"source":0,"locator":"createdAt"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/microsoft/rho-libero","title":"Hugging Face · microsoft/rho-libero","retrieved_at":"2026-10-13T08:54:57Z","sha256":"729eb280199c8d754a9a2395227da6234cfa4ca0c0ff63c073f47f8ce7edd9c0","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"microsoft/rho-libero","params_total":5247548896,"architecture":null,"context_length":null,"license":"mit","gated":false,"pipeline":null,"variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--microsoft--rho-libero","related":[],"org":{"slug":"microsoft","name":"Microsoft","kind":"empresa","url":"https://www.microsoft.com"}},{"id":"hf:Qwen/Qwen-Drive-1.0-4B","kind":"model_release","title":"Qwen publica Qwen-Drive-1.0-4B","date":"2026-08-27","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"Qwen/Qwen-Drive-1.0-4B","url":"https://huggingface.co/Qwen/Qwen-Drive-1.0-4B","entity":"qwen","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 4.5 B parámetros, licencia apache-2.0, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":4539265536,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"apache-2.0","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"image-text-to-text","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-27T08:14:27.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"QwenDriveForPlanning","evidence":{"source":1,"locator":"config.json#architectures[0]"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/Qwen/Qwen-Drive-1.0-4B","title":"Hugging Face · Qwen/Qwen-Drive-1.0-4B","retrieved_at":"2026-10-13T08:54:49Z","sha256":"d11958df008b21df92b2ef7519af525046d67b9e7f7697a398d0c9f3f60bbcc3","license":null},{"url":"https://huggingface.co/Qwen/Qwen-Drive-1.0-4B/blob/main/config.json","title":"Qwen/Qwen-Drive-1.0-4B · config.json","retrieved_at":"2026-10-13T08:54:49Z","sha256":"ffd92d5479d62cb8bb85a431139101b51d773ab27963addb84fa2137367daee0","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"Qwen/Qwen-Drive-1.0-4B","params_total":4539265536,"architecture":"QwenDriveForPlanning","context_length":null,"license":"apache-2.0","gated":false,"pipeline":"image-text-to-text","variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--qwen--qwen-drive-1.0-4b","related":[],"org":{"slug":"qwen","name":"Qwen (Alibaba)","kind":"laboratorio","url":"https://qwen.ai"}},{"id":"hf:allenai/Bwen-8B","kind":"model_release","title":"allenai publica Bwen-8B","date":"2026-08-26","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"allenai/Bwen-8B","url":"https://huggingface.co/allenai/Bwen-8B","entity":"allenai","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 8.3 B parámetros, licencia apache-2.0, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":8313169056,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"apache-2.0","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"text-generation","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-26T13:21:04.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"BolmoForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":32768,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/allenai/Bwen-8B","title":"Hugging Face · allenai/Bwen-8B","retrieved_at":"2026-10-13T08:55:12Z","sha256":"f4eba47e70dcba8e85e1ca72edddc8103f62a97378d3842fcf220355be1e8869","license":null},{"url":"https://huggingface.co/allenai/Bwen-8B/blob/main/config.json","title":"allenai/Bwen-8B · config.json","retrieved_at":"2026-10-13T08:55:12Z","sha256":"db088c1d87d3c29bd3c7b882490436daf40b2ff9312b8a47c4474fc6d876c4c9","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"allenai/Bwen-8B","params_total":8313169056,"architecture":"BolmoForCausalLM","context_length":32768,"license":"apache-2.0","gated":false,"pipeline":"text-generation","variants":[],"kv_bytes_per_token":147456,"quants":[]},"slug":"hf--allenai--bwen-8b","related":[],"org":{"slug":"allenai","name":"Allen Institute for AI","kind":"laboratorio","url":"https://allenai.org"}},{"id":"hf:allenai/Llama-3-Blama-8B","kind":"model_release","title":"allenai publica Llama-3-Blama-8B","date":"2026-08-26","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"allenai/Llama-3-Blama-8B","url":"https://huggingface.co/allenai/Llama-3-Blama-8B","entity":"allenai","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 8.3 B parámetros, licencia llama3, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":8250736800,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"llama3","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"text-generation","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-26T13:40:05.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"BolmoForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":8192,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/allenai/Llama-3-Blama-8B","title":"Hugging Face · allenai/Llama-3-Blama-8B","retrieved_at":"2026-10-13T08:55:11Z","sha256":"3a11acd80a53692772b4ef21bf71ca1e42e520bc192390256fcdd6c996bc40a2","license":null},{"url":"https://huggingface.co/allenai/Llama-3-Blama-8B/blob/main/config.json","title":"allenai/Llama-3-Blama-8B · config.json","retrieved_at":"2026-10-13T08:55:11Z","sha256":"77ecf00c8abafbe7430366cb5b14d241c2edf9a3cb082dec8ea11bc23a1d2fc6","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"allenai/Llama-3-Blama-8B","params_total":8250736800,"architecture":"BolmoForCausalLM","context_length":8192,"license":"llama3","gated":false,"pipeline":"text-generation","variants":[],"kv_bytes_per_token":131072,"quants":[]},"slug":"hf--allenai--llama-3-blama-8b","related":[],"org":{"slug":"allenai","name":"Allen Institute for AI","kind":"laboratorio","url":"https://allenai.org"}},{"id":"hf:inclusionAI/UI-Venus-2-9B","kind":"model_release","title":"inclusionAI publica UI-Venus-2-9B","date":"2026-08-26","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"inclusionAI/UI-Venus-2-9B","url":"https://huggingface.co/inclusionAI/UI-Venus-2-9B","entity":"inclusionai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Modelo nuevo: 0.0 B parámetros, 24 cuantizaciones GGUF disponibles."},"highlights":[],"facts":[{"field":"params_total","value":1469680,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"image-text-to-text","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-26T14:52:03.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"Qwen3_5ForConditionalGeneration","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}},{"field":"gguf_quants","value":24,"evidence":{"source":2,"locator":"siblings[].size"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/inclusionAI/UI-Venus-2-9B","title":"Hugging Face · inclusionAI/UI-Venus-2-9B","retrieved_at":"2026-10-13T08:55:36Z","sha256":"47fd493506be0de5ce7c7442ad4ba03b8dd1c6a3a49ea6c92b61b4e6b6d4dbe6","license":null},{"url":"https://huggingface.co/inclusionAI/UI-Venus-2-9B/blob/main/config.json","title":"inclusionAI/UI-Venus-2-9B · config.json","retrieved_at":"2026-10-13T08:55:37Z","sha256":"f6edd4ec68b2b9b46feb8b65ac165bdcb5971f96b1fdc2053d30e26625097ce1","license":null},{"url":"https://huggingface.co/bartowski/UI-Venus-2-9B-GGUF","title":"Hugging Face · bartowski/UI-Venus-2-9B-GGUF","retrieved_at":"2026-10-13T08:55:37Z","sha256":"83182d36bd3037a29c839efca33f5aedb397c4cfb7e40bbce4b848cd993d1cd7","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"inclusionAI/UI-Venus-2-9B","params_total":1469680,"architecture":"Qwen3_5ForConditionalGeneration","context_length":262144,"license":null,"gated":false,"pipeline":"image-text-to-text","variants":[],"kv_bytes_per_token":32768,"quants":[{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"IQ2_M","bytes":3769426720},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q2_K","bytes":4064273184},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"IQ3_XXS","bytes":4276020000},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"IQ3_XS","bytes":4560528160},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q3_K_S","bytes":4669317920},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"IQ3_M","bytes":4722533152},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q3_K_M","bytes":4920189728},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q2_K_L","bytes":5057553184},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q3_K_L","bytes":5111030560},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"IQ4_XS","bytes":5242643232},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"IQ4_NL","bytes":5478900512},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q4_0","bytes":5482832672},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q4_K_S","bytes":5598176032},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q4_K_M","bytes":5910782752},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q4_1","bytes":5944861472},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q3_K_XL","bytes":6001009440},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q5_K_S","bytes":6526427936},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q4_K_L","bytes":6665675552},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q5_K_M","bytes":6852928288},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q5_K_L","bytes":7480681248},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q6_K","bytes":7700259616},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q6_K_L","bytes":8192926496},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"Q8_0","bytes":9545982752},{"repo":"bartowski/UI-Venus-2-9B-GGUF","quant":"BF16","bytes":17920696800}]},"slug":"hf--inclusionai--ui-venus-2-9b","related":[],"org":{"slug":"inclusionai","name":"inclusionAI (Ant Group)","kind":"laboratorio","url":"https://github.com/inclusionAI"}},{"id":"hf:microsoft/rho-fr3-duo","kind":"model_release","title":"microsoft publica rho-fr3-duo","date":"2026-08-26","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"microsoft/rho-fr3-duo","url":"https://huggingface.co/microsoft/rho-fr3-duo","entity":"microsoft","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 5.2 B parámetros, licencia mit, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":5238067680,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"mit","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"created_at","value":"2026-08-26T21:09:30.000Z","evidence":{"source":0,"locator":"createdAt"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/microsoft/rho-fr3-duo","title":"Hugging Face · microsoft/rho-fr3-duo","retrieved_at":"2026-10-13T08:54:59Z","sha256":"b4dc79f7615d8d3a693238a43989f8663c5229d50caf36d92c1e4f940392367a","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"microsoft/rho-fr3-duo","params_total":5238067680,"architecture":null,"context_length":null,"license":"mit","gated":false,"pipeline":null,"variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--microsoft--rho-fr3-duo","related":[],"org":{"slug":"microsoft","name":"Microsoft","kind":"empresa","url":"https://www.microsoft.com"}},{"id":"hf:microsoft/rho-ur-ai-trainer","kind":"model_release","title":"microsoft publica rho-ur-ai-trainer","date":"2026-08-26","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"microsoft/rho-ur-ai-trainer","url":"https://huggingface.co/microsoft/rho-ur-ai-trainer","entity":"microsoft","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 5.2 B parámetros, licencia mit, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":5238067680,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"mit","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"created_at","value":"2026-08-26T21:10:08.000Z","evidence":{"source":0,"locator":"createdAt"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/microsoft/rho-ur-ai-trainer","title":"Hugging Face · microsoft/rho-ur-ai-trainer","retrieved_at":"2026-10-13T08:54:58Z","sha256":"0cc50f001a3bc136c62a9b2a6d4d1f5f511b7550ddd48f8ab3b41b98d22c95fa","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"microsoft/rho-ur-ai-trainer","params_total":5238067680,"architecture":null,"context_length":null,"license":"mit","gated":false,"pipeline":null,"variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--microsoft--rho-ur-ai-trainer","related":[],"org":{"slug":"microsoft","name":"Microsoft","kind":"empresa","url":"https://www.microsoft.com"}},{"id":"hf:microsoft/rho-yam-box","kind":"model_release","title":"microsoft publica rho-yam-box","date":"2026-08-26","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"microsoft/rho-yam-box","url":"https://huggingface.co/microsoft/rho-yam-box","entity":"microsoft","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 5.2 B parámetros, licencia mit, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":5238067680,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"mit","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"created_at","value":"2026-08-26T21:10:42.000Z","evidence":{"source":0,"locator":"createdAt"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/microsoft/rho-yam-box","title":"Hugging Face · microsoft/rho-yam-box","retrieved_at":"2026-10-13T08:54:58Z","sha256":"987c350a17a34c66031e985f62d965e9e76e1c889153039091eabe8a99b3e582","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"microsoft/rho-yam-box","params_total":5238067680,"architecture":null,"context_length":null,"license":"mit","gated":false,"pipeline":null,"variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--microsoft--rho-yam-box","related":[],"org":{"slug":"microsoft","name":"Microsoft","kind":"empresa","url":"https://www.microsoft.com"}},{"id":"hf:zai-org/GLM-5.3-BF16","kind":"model_release","title":"zai-org publica GLM-5.3-BF16","date":"2026-08-25","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"zai-org/GLM-5.3-BF16","url":"https://huggingface.co/zai-org/GLM-5.3-BF16","entity":"zai-org","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 753.3 B parámetros, licencia other:glm-5.3, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":753329940480,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"other:glm-5.3","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"text-generation","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-25T06:43:55.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"GlmMoeDsaForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":1048576,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/zai-org/GLM-5.3-BF16","title":"Hugging Face · zai-org/GLM-5.3-BF16","retrieved_at":"2026-10-13T08:55:07Z","sha256":"46ee66062cd71ec9c54e694f99fcb7bafb7b04dd48159b155eca7424b9bc0e1c","license":null},{"url":"https://huggingface.co/zai-org/GLM-5.3-BF16/blob/main/config.json","title":"zai-org/GLM-5.3-BF16 · config.json","retrieved_at":"2026-10-13T08:55:08Z","sha256":"186752be06e77dad34b363016211443822afeff1c12c0b63c8832fa60b61cafd","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"zai-org/GLM-5.3-BF16","params_total":753329940480,"architecture":"GlmMoeDsaForCausalLM","context_length":1048576,"license":"other:glm-5.3","gated":false,"pipeline":"text-generation","variants":[],"kv_bytes_per_token":89856,"quants":[]},"slug":"hf--zai-org--glm-5.3-bf16","related":[],"org":{"slug":"zai","name":"Z.ai (Zhipu)","kind":"laboratorio","url":"https://z.ai"}},{"id":"hf:zai-org/GLM-5.3-Flash-BF16","kind":"model_release","title":"zai-org publica GLM-5.3-Flash-BF16","date":"2026-08-25","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"zai-org/GLM-5.3-Flash-BF16","url":"https://huggingface.co/zai-org/GLM-5.3-Flash-BF16","entity":"zai-org","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Modelo nuevo: 321.3 B parámetros, licencia mit, 24 cuantizaciones GGUF disponibles."},"highlights":[],"facts":[{"field":"params_total","value":321323031390,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"mit","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"image-text-to-text","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-25T06:45:05.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"Glm5NextForConditionalGeneration","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":1048576,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}},{"field":"gguf_quants","value":24,"evidence":{"source":2,"locator":"siblings[].size"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/zai-org/GLM-5.3-Flash-BF16","title":"Hugging Face · zai-org/GLM-5.3-Flash-BF16","retrieved_at":"2026-10-13T08:55:07Z","sha256":"cc5094d5dbf7ef73565e8f1e56aa596c422fa0b6205bb2bb97be56f9ceeee58c","license":null},{"url":"https://huggingface.co/zai-org/GLM-5.3-Flash-BF16/blob/main/config.json","title":"zai-org/GLM-5.3-Flash-BF16 · config.json","retrieved_at":"2026-10-13T08:55:07Z","sha256":"3ea4bdc118c78b862c243aabc495aa1da1338d7844b23094de6b214d15735cb3","license":null},{"url":"https://huggingface.co/bartowski/GLM-5.3-Flash-BF16-GGUF","title":"Hugging Face · bartowski/GLM-5.3-Flash-BF16-GGUF","retrieved_at":"2026-10-13T08:55:07Z","sha256":"ee457bada24089bf99ba329bc4599f1435b88a625722ef48a0253ba24050a1c4","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"zai-org/GLM-5.3-Flash-BF16","params_total":321323031390,"architecture":"Glm5NextForConditionalGeneration","context_length":1048576,"license":"mit","gated":false,"pipeline":"image-text-to-text","variants":[],"kv_bytes_per_token":11264,"quants":[{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"IQ1_S","bytes":74845873152},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"IQ1_M","bytes":83014968416},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"IQ2_XXS","bytes":96630711424},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"IQ2_XS","bytes":101428274304},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"IQ2_S","bytes":107523515552},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"IQ2_M","bytes":120618067200},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"Q2_K","bytes":125671154912},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"IQ3_XXS","bytes":138825147648},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"IQ3_XS","bytes":147357148384},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"Q3_K_S","bytes":147357148384},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"Q3_K_M","bytes":155020141792},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"Q3_K_L","bytes":162437768544},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"IQ3_M","bytes":166627878240},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"IQ4_XS","bytes":177744881024},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"Q4_0","bytes":184093197696},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"Q4_K_S","bytes":189295994240},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"IQ4_NL","bytes":200828233184},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"Q4_K_M","bytes":200828233184},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"Q4_1","bytes":202609804736},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"Q5_K_S","bytes":227912951264},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"Q5_K_M","bytes":243691923040},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"Q6_K_S","bytes":268171978336},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"Q6_K","bytes":281044297408},{"repo":"bartowski/GLM-5.3-Flash-BF16-GGUF","quant":"Q8_0","bytes":341502043648}]},"slug":"hf--zai-org--glm-5.3-flash-bf16","related":[],"org":{"slug":"zai","name":"Z.ai (Zhipu)","kind":"laboratorio","url":"https://z.ai"}},{"id":"hf:zai-org/GLM-5.3-Flash","kind":"model_release","title":"zai-org publica GLM-5.3-Flash","date":"2026-08-25","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"zai-org/GLM-5.3-Flash","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","entity":"zai-org","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Modelo nuevo: 321.3 B parámetros, licencia mit, 12 cuantizaciones GGUF disponibles."},"highlights":[],"facts":[{"field":"params_total","value":321323031390,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"mit","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"image-text-to-text","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-25T06:43:14.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"Glm5NextForConditionalGeneration","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":1048576,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}},{"field":"gguf_quants","value":12,"evidence":{"source":2,"locator":"siblings[].size"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/zai-org/GLM-5.3-Flash","title":"Hugging Face · zai-org/GLM-5.3-Flash","retrieved_at":"2026-10-13T08:55:08Z","sha256":"6a2e22c6c88a94c659b18d96afbaf9cff77735b23489f94be580b911486759f0","license":null},{"url":"https://huggingface.co/zai-org/GLM-5.3-Flash/blob/main/config.json","title":"zai-org/GLM-5.3-Flash · config.json","retrieved_at":"2026-10-13T08:55:08Z","sha256":"e550dc05acbb033b359d2972047886b2f5bc1ff9b82f5752e420075bb4dfebf9","license":null},{"url":"https://huggingface.co/unsloth/GLM-5.3-Flash-GGUF","title":"Hugging Face · unsloth/GLM-5.3-Flash-GGUF","retrieved_at":"2026-10-13T08:55:09Z","sha256":"a7fadcfe79417e454ae8428b75f63c9a8dd6296186b054ddbeb590c956b9ae02","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"zai-org/GLM-5.3-Flash","params_total":321323031390,"architecture":"Glm5NextForConditionalGeneration","context_length":1048576,"license":"mit","gated":false,"pipeline":"image-text-to-text","variants":[],"kv_bytes_per_token":11264,"quants":[{"repo":"unsloth/GLM-5.3-Flash-GGUF","quant":"IQ1_S","bytes":93087245088},{"repo":"unsloth/GLM-5.3-Flash-GGUF","quant":"IQ1_M","bytes":97579344672},{"repo":"unsloth/GLM-5.3-Flash-GGUF","quant":"IQ2_XXS","bytes":101844951904},{"repo":"unsloth/GLM-5.3-Flash-GGUF","quant":"Q2_K_XL","bytes":108720071552},{"repo":"unsloth/GLM-5.3-Flash-GGUF","quant":"IQ3_XXS","bytes":120367571840},{"repo":"unsloth/GLM-5.3-Flash-GGUF","quant":"Q3_K_XL","bytes":147535922080},{"repo":"unsloth/GLM-5.3-Flash-GGUF","quant":"IQ4_XS","bytes":156822111200},{"repo":"unsloth/GLM-5.3-Flash-GGUF","quant":"Q4_K_XL","bytes":199707321472},{"repo":"unsloth/GLM-5.3-Flash-GGUF","quant":"Q5_K_XL","bytes":240306087008},{"repo":"unsloth/GLM-5.3-Flash-GGUF","quant":"Q6_K_XL","bytes":291833111808},{"repo":"unsloth/GLM-5.3-Flash-GGUF","quant":"Q8_0","bytes":340981966208},{"repo":"unsloth/GLM-5.3-Flash-GGUF","quant":"BF16","bytes":641641064288}]},"slug":"hf--zai-org--glm-5.3-flash","related":["github:ggml-org/llama.cpp:v0.6.0","github:sgl-project/sglang:v0.5.21"],"org":{"slug":"zai","name":"Z.ai (Zhipu)","kind":"laboratorio","url":"https://z.ai"}},{"id":"hf:zai-org/GLM-5.3","kind":"model_release","title":"zai-org publica GLM-5.3","date":"2026-08-25","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"zai-org/GLM-5.3","url":"https://huggingface.co/zai-org/GLM-5.3","entity":"zai-org","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Modelo nuevo: 753.3 B parámetros, licencia other:glm-5.3, 12 cuantizaciones GGUF disponibles."},"highlights":[],"facts":[{"field":"params_total","value":753329940480,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"other:glm-5.3","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"text-generation","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-25T06:42:50.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"GlmMoeDsaForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":1048576,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}},{"field":"gguf_quants","value":12,"evidence":{"source":2,"locator":"siblings[].size"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/zai-org/GLM-5.3","title":"Hugging Face · zai-org/GLM-5.3","retrieved_at":"2026-10-13T08:55:09Z","sha256":"c67c6a454fa040f11dd2e46ca97e8d83aafadba85ffcb4ec677f4a13c91adfc2","license":null},{"url":"https://huggingface.co/zai-org/GLM-5.3/blob/main/config.json","title":"zai-org/GLM-5.3 · config.json","retrieved_at":"2026-10-13T08:55:09Z","sha256":"346d7df7b4f78391b7801eb728dd20b2277736931c51d3b72882b300c43d1733","license":null},{"url":"https://huggingface.co/unsloth/GLM-5.3-GGUF","title":"Hugging Face · unsloth/GLM-5.3-GGUF","retrieved_at":"2026-10-13T08:55:09Z","sha256":"b3e2bb006fa4474c9c2516c4738bd2d95b06d0c0801286e77a227c580b1c8712","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"zai-org/GLM-5.3","params_total":753329940480,"architecture":"GlmMoeDsaForCausalLM","context_length":1048576,"license":"other:glm-5.3","gated":false,"pipeline":"text-generation","variants":[],"kv_bytes_per_token":89856,"quants":[{"repo":"unsloth/GLM-5.3-GGUF","quant":"IQ1_S","bytes":216715365893},{"repo":"unsloth/GLM-5.3-GGUF","quant":"IQ1_M","bytes":228492971557},{"repo":"unsloth/GLM-5.3-GGUF","quant":"IQ2_M","bytes":238577585701},{"repo":"unsloth/GLM-5.3-GGUF","quant":"Q2_K_XL","bytes":253878406789},{"repo":"unsloth/GLM-5.3-GGUF","quant":"IQ3_XXS","bytes":281688436357},{"repo":"unsloth/GLM-5.3-GGUF","quant":"Q3_K_XL","bytes":342965977029},{"repo":"unsloth/GLM-5.3-GGUF","quant":"IQ4_XS","bytes":365313228709},{"repo":"unsloth/GLM-5.3-GGUF","quant":"Q4_K_XL","bytes":467289116837},{"repo":"unsloth/GLM-5.3-GGUF","quant":"Q5_K_XL","bytes":562466263429},{"repo":"unsloth/GLM-5.3-GGUF","quant":"Q6_K_XL","bytes":684369515173},{"repo":"unsloth/GLM-5.3-GGUF","quant":"Q8_0","bytes":801357677216},{"repo":"unsloth/GLM-5.3-GGUF","quant":"BF16","bytes":1507988027936}]},"slug":"hf--zai-org--glm-5.3","related":["github:ggml-org/llama.cpp:v0.6.0","github:sgl-project/sglang:v0.5.21","rss:zai-api:9156c37fb1c2"],"org":{"slug":"zai","name":"Z.ai (Zhipu)","kind":"laboratorio","url":"https://z.ai"}},{"id":"hf:Qwen/Qwen3.8-Flash-Next","kind":"model_release","title":"Qwen publica Qwen3.8-Flash-Next","date":"2026-08-24","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"Qwen/Qwen3.8-Flash-Next","url":"https://huggingface.co/Qwen/Qwen3.8-Flash-Next","entity":"qwen","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Modelo nuevo: 180.0 B parámetros, licencia other:qwen-community-1.0, 12 cuantizaciones GGUF disponibles."},"highlights":[],"facts":[{"field":"params_total","value":179999981459,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"other:qwen-community-1.0","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"image-text-to-text","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-24T08:24:59.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"Qwen4ExpForConditionalGeneration","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}},{"field":"gguf_quants","value":12,"evidence":{"source":2,"locator":"siblings[].size"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/Qwen/Qwen3.8-Flash-Next","title":"Hugging Face · Qwen/Qwen3.8-Flash-Next","retrieved_at":"2026-10-13T08:54:50Z","sha256":"38ac1c8e554742aabfa645c72194f88cf816191e3bbf010cd6f94f6ef586cf9e","license":null},{"url":"https://huggingface.co/Qwen/Qwen3.8-Flash-Next/blob/main/config.json","title":"Qwen/Qwen3.8-Flash-Next · config.json","retrieved_at":"2026-10-13T08:54:50Z","sha256":"1c9f263b44653d8093f05e1e957efd9b49663785e33e5ec1ac0121f605eb5f79","license":null},{"url":"https://huggingface.co/unsloth/Qwen3.8-Flash-Next-GGUF","title":"Hugging Face · unsloth/Qwen3.8-Flash-Next-GGUF","retrieved_at":"2026-10-13T08:54:50Z","sha256":"954ac9c4571bc978993e3a1b37876e4e7030200dd92db25cd11abe3e7c88d04a","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"Qwen/Qwen3.8-Flash-Next","params_total":179999981459,"architecture":"Qwen4ExpForConditionalGeneration","context_length":262144,"license":"other:qwen-community-1.0","gated":false,"pipeline":"image-text-to-text","variants":[],"kv_bytes_per_token":24576,"quants":[{"repo":"unsloth/Qwen3.8-Flash-Next-GGUF","quant":"Q4_K_M","bytes":4693356736},{"repo":"unsloth/Qwen3.8-Flash-Next-GGUF","quant":"IQ1_S","bytes":72546461344},{"repo":"unsloth/Qwen3.8-Flash-Next-GGUF","quant":"IQ1_M","bytes":74538755776},{"repo":"unsloth/Qwen3.8-Flash-Next-GGUF","quant":"Q2_K_XL","bytes":78869128864},{"repo":"unsloth/Qwen3.8-Flash-Next-GGUF","quant":"IQ3_XXS","bytes":81961823936},{"repo":"unsloth/Qwen3.8-Flash-Next-GGUF","quant":"Q3_K_XL","bytes":89986353824},{"repo":"unsloth/Qwen3.8-Flash-Next-GGUF","quant":"IQ4_XS","bytes":93682584224},{"repo":"unsloth/Qwen3.8-Flash-Next-GGUF","quant":"Q4_K_XL","bytes":111334654784},{"repo":"unsloth/Qwen3.8-Flash-Next-GGUF","quant":"Q5_K_XL","bytes":158286406650},{"repo":"unsloth/Qwen3.8-Flash-Next-GGUF","quant":"Q6_K_XL","bytes":169165382688},{"repo":"unsloth/Qwen3.8-Flash-Next-GGUF","quant":"Q8_0","bytes":199286459744},{"repo":"unsloth/Qwen3.8-Flash-Next-GGUF","quant":"BF16","bytes":367028654272}]},"slug":"hf--qwen--qwen3.8-flash-next","related":["github:NVIDIA/TensorRT-LLM:v1.3.0rc27","github:NVIDIA/TensorRT-LLM:v1.3.0rc28","github:turboderp-org/exllamav3:v1.4.5","github:turboderp-org/exllamav3:v1.4.8","github:turboderp-org/exllamav3:v1.5.1"],"org":{"slug":"qwen","name":"Qwen (Alibaba)","kind":"laboratorio","url":"https://qwen.ai"}},{"id":"hf:inclusionAI/ArmorOCR","kind":"model_release","title":"inclusionAI publica ArmorOCR","date":"2026-08-20","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"inclusionAI/ArmorOCR","url":"https://huggingface.co/inclusionAI/ArmorOCR","entity":"inclusionai","channel":"n/a"},"status":"published","severity":"notable","action":"try","tags":["new-model"],"affects":["llama.cpp","lm-studio","ollama"],"summary":{"why":"Modelo nuevo: 8.8 B parámetros, licencia apache-2.0, 2 cuantizaciones GGUF disponibles."},"highlights":[],"facts":[{"field":"params_total","value":8767123696,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"apache-2.0","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"image-text-to-text","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-20T10:54:31.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"Qwen3VLForConditionalGeneration","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}},{"field":"gguf_quants","value":2,"evidence":{"source":2,"locator":"siblings[].size"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/inclusionAI/ArmorOCR","title":"Hugging Face · inclusionAI/ArmorOCR","retrieved_at":"2026-10-13T08:55:37Z","sha256":"c59799be88edeb682bf9042b18ecf9fa67c8dd5adfafae2ad1ecabab299e8082","license":null},{"url":"https://huggingface.co/inclusionAI/ArmorOCR/blob/main/config.json","title":"inclusionAI/ArmorOCR · config.json","retrieved_at":"2026-10-13T08:55:37Z","sha256":"85184045427d03a9efaad561fa4e52ce62da32078b5e671cddcb6a3d8115fe73","license":null},{"url":"https://huggingface.co/inclusionAI/ArmorOCR-GGUF","title":"Hugging Face · inclusionAI/ArmorOCR-GGUF","retrieved_at":"2026-10-13T08:55:38Z","sha256":"f3fe0c129040b162b1162caa538a8bc71c71fd5d6aa9c9e3acccb922c9392b62","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"inclusionAI/ArmorOCR","params_total":8767123696,"architecture":"Qwen3VLForConditionalGeneration","context_length":262144,"license":"apache-2.0","gated":false,"pipeline":"image-text-to-text","variants":[],"kv_bytes_per_token":147456,"quants":[{"repo":"inclusionAI/ArmorOCR-GGUF","quant":"Q4_K_M","bytes":5027784928},{"repo":"inclusionAI/ArmorOCR-GGUF","quant":"Q8_0","bytes":8709519584}]},"slug":"hf--inclusionai--armorocr","related":[],"org":{"slug":"inclusionai","name":"inclusionAI (Ant Group)","kind":"laboratorio","url":"https://github.com/inclusionAI"}},{"id":"hf:ibm-granite/granite-speech-5.0-470m-turboctc-nc","kind":"model_release","title":"ibm-granite publica granite-speech-5.0-470m-turboctc-nc","date":"2026-08-17","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"ibm-granite/granite-speech-5.0-470m-turboctc-nc","url":"https://huggingface.co/ibm-granite/granite-speech-5.0-470m-turboctc-nc","entity":"ibm-granite","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 0.5 B parámetros, licencia cc-by-nc-sa-4.0, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":472993792,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"cc-by-nc-sa-4.0","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"automatic-speech-recognition","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-17T17:29:48.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"GraniteSpeech5ForCTC","evidence":{"source":1,"locator":"config.json#architectures[0]"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/ibm-granite/granite-speech-5.0-470m-turboctc-nc","title":"Hugging Face · ibm-granite/granite-speech-5.0-470m-turboctc-nc","retrieved_at":"2026-10-13T08:55:10Z","sha256":"c7db78d261b6a2d0f8035afefaecd51e06982f81d36010efcfd2b154e01c9ccb","license":null},{"url":"https://huggingface.co/ibm-granite/granite-speech-5.0-470m-turboctc-nc/blob/main/config.json","title":"ibm-granite/granite-speech-5.0-470m-turboctc-nc · config.json","retrieved_at":"2026-10-13T08:55:10Z","sha256":"546d31e0e2f663bbe21397de9c555c95602f2435b41f4b379235dc8a0973325d","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"ibm-granite/granite-speech-5.0-470m-turboctc-nc","params_total":472993792,"architecture":"GraniteSpeech5ForCTC","context_length":null,"license":"cc-by-nc-sa-4.0","gated":false,"pipeline":"automatic-speech-recognition","variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--ibm-granite--granite-speech-5.0-470m-turboctc-nc","related":[],"org":{"slug":"ibm","name":"IBM","kind":"empresa","url":"https://www.ibm.com"}},{"id":"hf:nvidia/NVIDIA-Nemotron-Labs-Teacher-Chat","kind":"model_release","title":"nvidia publica NVIDIA-Nemotron-Labs-Teacher-Chat","date":"2026-08-14","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"nvidia/NVIDIA-Nemotron-Labs-Teacher-Chat","url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-Chat","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 560.5 B parámetros, licencia other:openmdw-1.1, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":560524578816,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"other:openmdw-1.1","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"text-generation","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-14T17:09:11.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"NemotronHForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-Chat","title":"Hugging Face · nvidia/NVIDIA-Nemotron-Labs-Teacher-Chat","retrieved_at":"2026-10-13T08:55:03Z","sha256":"5a9383ae8d14955238115abd8d2057f4598e14642eb36cccf42f223b723f2270","license":null},{"url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-Chat/blob/main/config.json","title":"nvidia/NVIDIA-Nemotron-Labs-Teacher-Chat · config.json","retrieved_at":"2026-10-13T08:55:04Z","sha256":"1c3dc1e686e1f24d5a11c9efced86f9f988f63e2637b685ed70668006d82ed71","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"nvidia/NVIDIA-Nemotron-Labs-Teacher-Chat","params_total":560524578816,"architecture":"NemotronHForCausalLM","context_length":262144,"license":"other:openmdw-1.1","gated":false,"pipeline":"text-generation","variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--nvidia--nvidia-nemotron-labs-teacher-chat","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}},{"id":"hf:nvidia/NVIDIA-Nemotron-Labs-Teacher-Competition-Coding","kind":"model_release","title":"nvidia publica NVIDIA-Nemotron-Labs-Teacher-Competition-Coding","date":"2026-08-14","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"nvidia/NVIDIA-Nemotron-Labs-Teacher-Competition-Coding","url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-Competition-Coding","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 560.5 B parámetros, licencia other:openmdw-1.1, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":560524578816,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"other:openmdw-1.1","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"text-generation","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-14T17:08:58.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"NemotronHForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-Competition-Coding","title":"Hugging Face · nvidia/NVIDIA-Nemotron-Labs-Teacher-Competition-Coding","retrieved_at":"2026-10-13T08:55:04Z","sha256":"54851b347462f78ce05b6cb6eb92db09f3951e83c93827a088b27a6a81d3ebb8","license":null},{"url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-Competition-Coding/blob/main/config.json","title":"nvidia/NVIDIA-Nemotron-Labs-Teacher-Competition-Coding · config.json","retrieved_at":"2026-10-13T08:55:04Z","sha256":"1c3dc1e686e1f24d5a11c9efced86f9f988f63e2637b685ed70668006d82ed71","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"nvidia/NVIDIA-Nemotron-Labs-Teacher-Competition-Coding","params_total":560524578816,"architecture":"NemotronHForCausalLM","context_length":262144,"license":"other:openmdw-1.1","gated":false,"pipeline":"text-generation","variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--nvidia--nvidia-nemotron-labs-teacher-competition-coding","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}},{"id":"hf:nvidia/NVIDIA-Nemotron-Labs-Teacher-General-Reasoning","kind":"model_release","title":"nvidia publica NVIDIA-Nemotron-Labs-Teacher-General-Reasoning","date":"2026-08-14","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"nvidia/NVIDIA-Nemotron-Labs-Teacher-General-Reasoning","url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-General-Reasoning","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 560.5 B parámetros, licencia other:openmdw-1.1, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":560524578816,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"other:openmdw-1.1","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"text-generation","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-14T17:08:25.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"NemotronHForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-General-Reasoning","title":"Hugging Face · nvidia/NVIDIA-Nemotron-Labs-Teacher-General-Reasoning","retrieved_at":"2026-10-13T08:55:05Z","sha256":"5ba366d0b410f02d0796d211ce92ec974fa723027706bb67efa382af0de1c55e","license":null},{"url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-General-Reasoning/blob/main/config.json","title":"nvidia/NVIDIA-Nemotron-Labs-Teacher-General-Reasoning · config.json","retrieved_at":"2026-10-13T08:55:05Z","sha256":"1c3dc1e686e1f24d5a11c9efced86f9f988f63e2637b685ed70668006d82ed71","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"nvidia/NVIDIA-Nemotron-Labs-Teacher-General-Reasoning","params_total":560524578816,"architecture":"NemotronHForCausalLM","context_length":262144,"license":"other:openmdw-1.1","gated":false,"pipeline":"text-generation","variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--nvidia--nvidia-nemotron-labs-teacher-general-reasoning","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}},{"id":"hf:nvidia/NVIDIA-Nemotron-Labs-Teacher-Instruction-Following","kind":"model_release","title":"nvidia publica NVIDIA-Nemotron-Labs-Teacher-Instruction-Following","date":"2026-08-14","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"nvidia/NVIDIA-Nemotron-Labs-Teacher-Instruction-Following","url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-Instruction-Following","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 560.5 B parámetros, licencia other:openmdw-1.1, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":560524578816,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"other:openmdw-1.1","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"text-generation","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-14T17:08:45.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"NemotronHForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-Instruction-Following","title":"Hugging Face · nvidia/NVIDIA-Nemotron-Labs-Teacher-Instruction-Following","retrieved_at":"2026-10-13T08:55:05Z","sha256":"81a2787e80f73bfeb52d9763de0bf255a10ce6accdc16bc59b68bde65ca04492","license":null},{"url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-Instruction-Following/blob/main/config.json","title":"nvidia/NVIDIA-Nemotron-Labs-Teacher-Instruction-Following · config.json","retrieved_at":"2026-10-13T08:55:05Z","sha256":"1c3dc1e686e1f24d5a11c9efced86f9f988f63e2637b685ed70668006d82ed71","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"nvidia/NVIDIA-Nemotron-Labs-Teacher-Instruction-Following","params_total":560524578816,"architecture":"NemotronHForCausalLM","context_length":262144,"license":"other:openmdw-1.1","gated":false,"pipeline":"text-generation","variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--nvidia--nvidia-nemotron-labs-teacher-instruction-following","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}},{"id":"hf:nvidia/NVIDIA-Nemotron-Labs-Teacher-STEM","kind":"model_release","title":"nvidia publica NVIDIA-Nemotron-Labs-Teacher-STEM","date":"2026-08-14","observed_at":"2026-10-13T08:55:38Z","subject":{"type":"model","ref":"nvidia/NVIDIA-Nemotron-Labs-Teacher-STEM","url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-STEM","entity":"nvidia","channel":"n/a"},"status":"published","severity":"notable","action":"watch","tags":["new-model"],"affects":[],"summary":{"why":"Modelo nuevo: 560.5 B parámetros, licencia other:openmdw-1.1, aún sin GGUF de referencia."},"highlights":[],"facts":[{"field":"params_total","value":560524578816,"evidence":{"source":0,"locator":"safetensors.total"}},{"field":"license","value":"other:openmdw-1.1","evidence":{"source":0,"locator":"cardData.license"}},{"field":"gated","value":false,"evidence":{"source":0,"locator":"gated"}},{"field":"pipeline","value":"text-generation","evidence":{"source":0,"locator":"pipeline_tag"}},{"field":"created_at","value":"2026-08-14T17:07:48.000Z","evidence":{"source":0,"locator":"createdAt"}},{"field":"architecture","value":"NemotronHForCausalLM","evidence":{"source":1,"locator":"config.json#architectures[0]"}},{"field":"context_length","value":262144,"evidence":{"source":1,"locator":"config.json#max_position_embeddings"}}],"verification":{"status":"verified","checked_at":"2026-10-13T08:55:38Z","rejected":[]},"triage":{"method":"rules","model":null,"reasons":["modelo nuevo de organización vigilada; hechos desde la API del Hub"]},"sources":[{"url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-STEM","title":"Hugging Face · nvidia/NVIDIA-Nemotron-Labs-Teacher-STEM","retrieved_at":"2026-10-13T08:55:06Z","sha256":"172588c4c56b90128dec0ba127fe1b0c59bca6c00234e0834bbafd6efa654c06","license":null},{"url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Teacher-STEM/blob/main/config.json","title":"nvidia/NVIDIA-Nemotron-Labs-Teacher-STEM · config.json","retrieved_at":"2026-10-13T08:55:06Z","sha256":"1c3dc1e686e1f24d5a11c9efced86f9f988f63e2637b685ed70668006d82ed71","license":null}],"attribution":{"data_license":"CC-BY-4.0","notes":null},"model":{"hf_id":"nvidia/NVIDIA-Nemotron-Labs-Teacher-STEM","params_total":560524578816,"architecture":"NemotronHForCausalLM","context_length":262144,"license":"other:openmdw-1.1","gated":false,"pipeline":"text-generation","variants":[],"kv_bytes_per_token":null,"quants":[]},"slug":"hf--nvidia--nvidia-nemotron-labs-teacher-stem","related":[],"org":{"slug":"nvidia","name":"NVIDIA","kind":"empresa","url":"https://developer.nvidia.com"}}]}