DeepSeek released dsh-superagent, an MIT-licensed research prototype that routes GitHub and Feishu requests to parallel worker sessions (MarkTechPost). In simulation, the system reduced p95 task completion time from 29.5 minutes to 15.7 minutes. Separately, Nace AI open-sourced Drex Max v1, a 27B multimodal decision model that runs on Apple Silicon and scores 67.52 (MarkTechPost).
OpenAI announced that starting in December, its Plus, Pro, and Business plans will shift from unlimited-prompt pricing to a shared token-based quota system that covers both Chat and Work modes (The Decoder). More complex models will consume higher quota amounts, while simple instant chats will not count toward the limit. Anthropic expanded its text watermark to four additional Claude models to meet EU AI Act requirements by December, though the watermark contains a bypass on older Claude Opus 4.6 and Sonnet 4.6 models via a sampling parameter adjustment (The Decoder).