{"items":[{"id":"072d297d","date":"2026-11-10T14:02:39.982Z","title":"Nvidia releases GR00T 2 world-action models for robots","url":"https://research.nvidia.com/labs/gear/gr00t-2","note":"A diffusion transformer that predicts future video and actions together learns skills from human video: 40% zero-shot on 50 unseen tasks, 95% after five demos and a 30-minute fine-tune on a single RTX 5090.","tag":"model"},{"id":"770d0cf6","date":"2026-11-10T14:02:39.583Z","title":"Anthropic releases Claude Fable 5.2 and Mythos 5.2 at lower prices","url":"https://www.anthropic.com/claude-fable-and-mythos-5-2","note":"The frontier Fable/Mythos tier drops to $8/$40 per million tokens with outputs up to 30% faster than 5.1, plus an optional fast mode at 2.5x generation speed for $16/$80.","tag":"news"},{"id":"1fd9a3ec","date":"2026-11-10T14:02:39.197Z","title":"Meituan releases LongCat-2.5, a 1.6T MoE under MIT with native vision","url":"https://huggingface.co/meituan-longcat/LongCat-2.5","note":"1.6T total parameters with roughly 48B active per token, MIT-licensed with native visual understanding, weights on Hugging Face a week after the hosted API launch — a serious open option for long-horizon agentic coding.","tag":"model"},{"id":"d2f6369c","date":"2026-11-10T10:19:40.575Z","title":"Nous Research releases Arete-27B, an Apache 2.0 value fine-tune of Qwen3.8-27B","url":"https://huggingface.co/NousResearch/Arete-27B","note":"Nous post-trained Qwen3.8-27B under Apache 2.0 to remove state-aligned refusals: propaganda-style answers dropped from 22% to 0.9% with no capability regression — a drop-in swap for anyone running Qwen3.8-27B locally.","tag":"model"},{"id":"5951d1ab","date":"2026-11-10T10:19:38.723Z","title":"Thinking Machines releases Inkling 2 Small, a 594B open-weight MoE","url":"https://thinkingmachines.ai/news/inkling-2-small/","note":"Thinking Machines second-generation open-weight model: 594B total but only 17B active parameters, image inputs and variable effort, scoring 31 on the Artificial Analysis Intelligence Index.","tag":"model"},{"id":"b5e14022","date":"2026-11-10T10:19:36.997Z","title":"Kimi K3.1 Pro released as open-weight 3T-class multimodal model","url":"https://huggingface.co/moonshotai/Kimi-K3.1-Pro","note":"Moonshot open-weighted its 3T-class multimodal agentic model with a 1M-token context and big gains in agentic coding — frontier weights you can inspect and self-host, albeit at extreme scale.","tag":"model"}]}