Meituan publishes weights for its 1.6-trillion-parameter LongCat-2.5
The MIT-licensed agent model trails Claude Opus 5 on Meituan's harder terminal tests, and its self-hosting guide covers only Huawei Ascend chips.
Open source, from The Decoder
A day of caveats attached to capability. Meituan's open model trails on harder tests and ships hosting guidance for one chip family, while Anthropic and Uber both describe layers built to check agent work, by risk or by context. The painting demos come from staff, with no availability stated.
4 stories, about 6 minutes to read them all.
The MIT-licensed agent model trails Claude Opus 5 on Meituan's harder terminal tests, and its self-hosting guide covers only Huawei Ascend chips.
Open source, from The Decoder
Anthropic says merged pull requests to the Claude Code repo more than doubled between June and September while reverts fell by over 40%.
Coding tools, from InfoQ
Uber says identity and API controls cannot judge fit, so it added registry, guard and replay layers across the agent lifecycle.
Safety and security, from InfoQ
The model writes Python that places simulated brushstrokes, one scene using roughly 700,000. The demos come from Anthropic staff, and availability is not stated.
Models, from The Decoder