https://openmodelweights.com/news/nvidia-cosmos3-h-dreams-open-surgical-world-model-20261021/Open Model Weightsen2026-10-21T06:00:56ZNVIDIA releases Cosmos3-H-Dreams world model for surgical robotics https://openmodelweights.com/news/google-torchtpu-pytorch-open-models-native-tpu-20261021/Open Model Weightsen2026-10-21T06:00:56ZGoogle open-sources TorchTPU for native PyTorch and open-weight AI https://openmodelweights.com/news/pytorch-215-apple-mps-fp8-local-llm-performance-20261021/Open Model Weightsen2026-10-21T06:00:56ZPyTorch 2.15 accelerates local open-weight AI on Apple Silicon https://openmodelweights.com/news/anylanguagemodel-10-swift-local-open-weight-apple-20261021/Open Model Weightsen2026-10-21T06:00:56ZAnyLanguageModel 1.0 gives Swift apps one API for local open weights https://openmodelweights.com/news/nvidia-nemotron35-parse-open-document-ocr-layout-20261020/Open Model Weightsen2026-10-20T07:06:00ZNVIDIA opens Nemotron 3.5 Parse for document OCR, charts and structured layouts https://openmodelweights.com/news/huggingface-torchneuron-trainium-native-pytorch-20261020/Open Model Weightsen2026-10-20T07:06:00ZHugging Face brings native PyTorch open-model workflows to AWS Trainium with TorchNeuron https://openmodelweights.com/news/lighton-slideon-mslideon-sparse-semantic-retrieval-20261020/Open Model Weightsen2026-10-20T07:06:00ZLightOn releases SlideOn and mSlideOn open sparse retrieval models for RAG https://openmodelweights.com/news/qvac-transcribepsy-254m-local-multilingual-asr-20261020/Open Model Weightsen2026-10-20T07:06:00ZQVAC releases 254M TranscribePsy open weights for on-device speech recognition https://openmodelweights.com/news/inco-dflash2-qwen38-flash-next-speculative-decoding-20261019/Open Model Weightsen2026-10-19T16:37:02ZInco AI releases DFlash 2 draft weights for Qwen3.8 Flash Next https://openmodelweights.com/news/tencent-hy-memory-30b-a3b-open-agent-memory-writer-20261019/Open Model Weightsen2026-10-19T16:37:02ZTencent opens Hy-Memory 30B-A3B for persistent AI agent memory writing https://openmodelweights.com/news/mistral-timestral-209m-zero-shot-forecasting-20261019/Open Model Weightsen2026-10-19T16:37:02ZMistral releases Timestral open-weight forecasting models from 5M to 209M https://openmodelweights.com/news/sglang-0522-instanttensor-multimodal-open-weight-serving-20261019/Open Model Weightsen2026-10-19T09:28:07ZSGLang 0.5.22 adds faster model loading, more open-weight formats and an important state fix https://openmodelweights.com/news/goto-ai-gemma3-27b-indonesian-local-languages-20261019/Open Model Weightsen2026-10-19T09:28:07ZGoTo releases a 27B Gemma-based open-weight model for Indonesian and regional languages https://openmodelweights.com/news/stepfun-step5-preview-600b-27b-weights-license-20261019/Open Model Weightsen2026-10-19T09:28:07ZStepFun opens Step 5 Preview's 600B MoE weights with a custom commercial-use license https://openmodelweights.com/news/nvidia-nemotron35-bolt-24gb-local-vision-agents-20261019/Open Model Weightsen2026-10-19T09:28:07ZNVIDIA releases Nemotron 3.5 Bolt as a 24GB local vision-language agent model https://openmodelweights.com/news/nvidia-nemotron35-super-vl-120b-multimodal-20261019/Open Model Weightsen2026-10-19T09:28:07ZNVIDIA opens Nemotron 3.5 Super VL, a 120B multimodal model for enterprise agents