Meituan has released LongCat-2.5, a 1.6-trillion-parameter multimodal model with a 1 million token context window. The model is designed for long-horizon agent tasks and was trained on Chinese chips. Its API is compatible with Anthropic's endpoint, and Meituan claims its performance benchmarks fall between Claude Opus 4.8 and Claude Opus 5, even surpassing Opus 5 on some metrics. While weights are not yet released, Meituan has provided a guide for integrating LongCat-2.5 into Claude Code, though this integration currently results in a 200K token context window due to Claude Code's auto-compaction for unrecognized models. AI
IMPACT Meituan's LongCat-2.5 release challenges existing models with its large context window and multimodal capabilities, potentially influencing agent development.
RANK_REASON Frontier-lab model release with system card [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- Anthropic
- Claude Code
- Claude Opus
- GPT-5.5
- GPT 5.6 "Sol"
- Hacker News
- Hugging Face
- LongCat-2.0
- LongCat-2.5
- Meituan
- MIT License
- Opus 4.8
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →