https://openmodelweights.com/news/tencent-hy-simulmt-open-weight-live-speech-translation-20261028/Open Model Weightsen2026-10-28T20:39:52ZTencent opens Hy-SimulMT for streaming speech translation https://openmodelweights.com/news/vllm-semantic-router-d3-ultra-open-multimodal-decisions-20261028/Open Model Weightsen2026-10-28T20:39:52ZvLLM Semantic Router releases d3-ultra for multimodal decisions https://openmodelweights.com/news/huggingface-diffusers-0421-remote-code-safety-fix-20261028/Open Model Weightsen2026-10-28T20:39:52ZDiffusers 0.42.1 closes remote-code bypass in model loading https://openmodelweights.com/news/llamacpp-lfm3-local-text-vision-audio-support-20261028/Open Model Weightsen2026-10-28T20:39:52Zllama.cpp adds local LFM3 text, vision and audio model support https://openmodelweights.com/news/liquid-ai-lfm3-six-multimodal-models-edge-20261028/Open Model Weightsen2026-10-28T09:52:08ZLiquid AI releases LFM3: six compact models share one backbone https://openmodelweights.com/news/baidu-unlimited-ocr-2-long-document-open-weights-20261028/Open Model Weightsen2026-10-28T09:52:08ZBaidu releases Unlimited OCR 2 for longer document parsing https://openmodelweights.com/news/decisionformer-open-decision-model-training-framework-20261028/Open Model Weightsen2026-10-28T09:52:08ZDecisionFormer opens one-pass decision-model training for agents https://openmodelweights.com/news/nanbeige43-33b-open-agentic-teacher-model-20261027/Open Model Weightsen2026-10-27T07:24:38ZNanbeige opens a 33B agent model built to teach smaller models https://openmodelweights.com/news/qwen-audio-31-asr-speaker-diarization-20261027/Open Model Weightsen2026-10-27T07:24:38ZQwen extends speech recognition to speaker-aware audio analysis https://openmodelweights.com/news/sglang-unified-memory-hybrid-cache-20261027/Open Model Weightsen2026-10-27T07:24:38ZSGLang shares GPU memory across KV cache and recurrent states https://openmodelweights.com/news/llamacpp-deepseek-v41-flash-local-support-20261027/Open Model Weightsen2026-10-27T07:24:38Zllama.cpp adds a DeepSeek V4.1 Flash model path for local runtimes