AI Briefing#open-weights
The latest AI and developer news about #open-weights, with the original source and a short summary.
The latest AI and developer news about #open-weights, with the original source and a short summary.
Cohere has released Command A+, an open-weight model based on Apache 2.0.
Stability AI has released Stable Audio 3.0, an open-weight audio model trained on licensed data.
Stability AI released the Stable Audio 3.0 model family, trained on fully licensed data.
Qwen3.5-9B's political censorship was found to be a small circuit that decomposes into 3 axes.
Hugging Face has released a leaderboard that compares agent systems' performance and cost together.
MiroThinker-1.7, an open-weight deep research agent based on Qwen3 MoE, has been released.
Datadog released Toto 2.0, ranging from 4m to 2.5B parameters, demonstrating clear scaling effects.
DeepSeek V4 Pro scored a middling 77, while Flash's API flaw surfaced even at just 2 cents.
Mistral AI is rapidly expanding into the enterprise market with a sovereign open-weight strategy.
Analysis suggests that the reduced openness and tightened licensing of open-weight models could lead to monopolization and concentrated pricing power in the AI market.
Artificial Analysis ranked the anonymous model Peanut #8 on its Text-to-Image leaderboard.
Moonshot AI's Kimi K2.6 won an AI coding challenge, edging out GPT-5.5 and Claude.
Mistral is differentiating itself from OpenAI by targeting European enterprise and public-sector markets through data sovereignty and an open-weight strategy.
Open-weight Kimi K2.6 outperformed Claude and GPT-5.5 in an AI coding challenge.
Mistral has unveiled Medium 3.5, a 128B model specialized in coding and reasoning, and Vibe, a cloud-based coding agent.
On the Arena leaderboard, Xiaomi's MIT-licensed model Mimo-v2.5 Pro surpassed Claude Opus 4.5.
Poolside has unveiled the Laguna XS.2 and Laguna M.1 models, optimized for coding agent performance.
On Terminal-Bench 2.0, 27B-32B open-weight coding models recorded up to 38.2%.
This analyzes why AI application companies should adopt post-training based on open-weight models to cut costs and optimize performance.
Ling-2.6-1T is released as Open Weights, and the same policy applies to the 104B Flash as well.
OpenAI released the Privacy Filter open-weight model under Apache 2.0.
A new clause has been added to the MiniMax-M2.7 model license permitting free use for personal and non-profit research purposes.