Meituan officially releases LongCat-2.0: a trillion-parameter large model trained on domestic compute clusters
Original title:美团 LongCat-2.0 正式发布:在国产算力集群上完成全流程训练与推理的万亿参数模型
AI summary
Meituan released and open-sourced its new-generation trillion-parameter large model LongCat-2.0 on June 30. It has 1.6T total parameters, about 48B average activated parameters, native support for a 1M ultra-long context, and completed full-process training and inference on a 50,000-GPU domestic compute cluster. It adopts LSA sparse attention, zero-compute experts, ScMoE and MOPD multi-expert fusion (Agent/Reasoning/Interaction expert groups). In evaluations, it scored 59.5 on SWE-bench Pro and 77.3 on SWE-bench Multilingual. The preview is available via OpenRouter and longcat.ai, and its monthly call volume ranks among OpenRouter's global top three.
Key Takeaways
- Meituan released and open-sourced its new-generation trillion-parameter large model LongCat-2.
- 0 on June 30.
- It has 1.
- 6T total parameters,about 48B average activated parameters,native support for a 1M ultra-long…
- It adopts LSA sparse attention,zero-compute experts,ScMoE and MOPD multi-expert fusion…
Quick Facts
- Published
- Updated
- Source
- 公众号:龙猫LongCat(美团)
- Category
- Model Releases
- AI score
- 82
国产算力上首个全流程自训的万亿开源模型,1M上下文和动态专家架构直指Agentic Coding场景,OpenRouter调用量已经冲到前三,不是Demo是生产力。
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Meituan officially releases LongCat-2.0: a trillion-parameter large model trained on domestic compute clusters. July 1, 2026. https://new.nat6.net/en/news/cmr1js18u03imslnld2yrjboo/