LongCat-Flash-Thinking-2601
Update to the 560B thinking model for agentic work, with environment scaling, multi-environment RL, tolerance for noisy tools and a Heavy Thinking parallel mode.
Trains under injected environment noise to generalize to real, imperfect tool settings; adds Heavy Thinking mode (intensive parallel reasoning) for hard tasks. MIT.
- Date
- Wednesday, 14 January 2026
- Lab
- Meituan (LongCat)
- Kind
- open-weights
- Access
- open weights
Date from HF initial commit; no benchmark numbers recorded.
Sources
This record was checked against its sources on 6 October 2026. How we check