https://openmodelweights.com/news/Daily source-first intelligence on open-weight AI.enhttps://openmodelweights.com/news/ai2-astabrief-8b-open-weight-scientific-reports/https://openmodelweights.com/news/ai2-astabrief-8b-open-weight-scientific-reports/Fri, 02 Oct 2026 15:00:00 +0000Model releaseAstaBrief turns retrieved scientific evidence into cited reports and gives institutions an open-weight option for running the report-writing stage on their own infrastructure.Ai2 / Hugging Facehttps://openmodelweights.com/news/llama-cpp-adds-decision-models-system-one-endpoint/https://openmodelweights.com/news/llama-cpp-adds-decision-models-system-one-endpoint/Fri, 02 Oct 2026 14:30:00 +0000RuntimeThe local inference stack can now score typed options directly instead of generating free-form text, opening a different deployment path for routing, moderation and agent control.ggml-org / Hugging Facehttps://openmodelweights.com/news/olmo-core-3-open-moe-training-stack/https://openmodelweights.com/news/olmo-core-3-open-moe-training-stack/Fri, 02 Oct 2026 13:40:00 +0000Training infrastructureAi2 redesigned its open training infrastructure around MoE routing, expert parallelism and lower-precision execution, with the next generation of Olmo expected to use the stack.Ai2 / Hugging Facehttps://openmodelweights.com/news/nvidia-kumo-tabular-open-foundation-model/https://openmodelweights.com/news/nvidia-kumo-tabular-open-foundation-model/Fri, 02 Oct 2026 12:50:00 +0000Model releaseThe Kumo Structured family targets classification and regression directly from labeled tables, with three small model sizes and a commercial-use license.NVIDIA / Hugging Facehttps://openmodelweights.com/news/holo4-open-agent-models-27b-35b/https://openmodelweights.com/news/holo4-open-agent-models-27b-35b/Fri, 02 Oct 2026 12:10:00 +0000Model releaseH Company’s new agentic family is designed to move between GUIs, code, MCP and APIs, with downloadable weights in multiple precisions and public benchmark trajectories.H Company / Hugging Facehttps://openmodelweights.com/news/liquid-ai-lfm25-vl-dspark-speculative-decoding/https://openmodelweights.com/news/liquid-ai-lfm25-vl-dspark-speculative-decoding/Fri, 02 Oct 2026 11:30:00 +0000InferenceThe DSpark companion model adds speculative decoding to Liquid AI’s vision-language stack with integrations for llama.cpp, MLX-VLM and SGLang.Liquid AI / Hugging Facehttps://openmodelweights.com/news/flux-3-action-open-weight-world-action-model/https://openmodelweights.com/news/flux-3-action-open-weight-world-action-model/Fri, 02 Oct 2026 10:55:00 +0000Model releaseFLUX 3 Action predicts future frames and actions together and arrives with robot-policy checkpoints, LeRobot integration and a dedicated weights license.Black Forest Labs / Hugging Facehttps://openmodelweights.com/news/nvidia-nemotron-3-diarization-open-weight-100m/https://openmodelweights.com/news/nvidia-nemotron-3-diarization-open-weight-100m/Fri, 02 Oct 2026 10:20:00 +0000Model releaseNemotron 3 Diarization targets real-time and offline speaker attribution, turning speech streams into speaker-aware timelines for meetings, calls and voice agents.NVIDIA / Hugging Facehttps://openmodelweights.com/news/transformers-runs-gguf-llama-cpp-quants/https://openmodelweights.com/news/transformers-runs-gguf-llama-cpp-quants/Fri, 02 Oct 2026 09:40:00 +0000RuntimeGGUF checkpoints can now be loaded through familiar Transformers APIs while reusing ggml kernels to target practical local inference, initially with an Apple Silicon focus.Hugging Facehttps://openmodelweights.com/news/nvidia-agrees-to-acquire-hugging-face-open-platform-pledge/https://openmodelweights.com/news/nvidia-agrees-to-acquire-hugging-face-open-platform-pledge/Fri, 02 Oct 2026 09:00:00 +0000EcosystemThe $12.93 billion agreement would put the central model-sharing platform inside NVIDIA while the companies publicly commit to model, cloud and accelerator choice.NVIDIA