Meituan Open-Sources Trillion-Parameter Model LongCat-2.0
智东西 · wechat · 2026-07-06
Meituan open-sources its trillion-parameter large model LongCat-2.0, with 1.6 trillion total parameters and 48 billion average activated parameters. It is the industry's first trillion-parameter model to complete full training and inference on a 50,000-card domestic computing cluster, and releases inference code optimized for domestic chips. Scored 59.5 on SWE-benchPro, surpassing GPT-5.5 (58.6) and Claude Opus 4.6 (57.3). Uses MoE architecture, natively supports 1M token context, introduces LongCat Sparse Attention (LSA) to reduce quadratic complexity to linear, and adds N-gram Embedding parameter expansion and multi-teacher online distillation.
Related event: Meituan Open-Sources LongCat-2.0 on Domestic Chips(3 posts)→
More from Infra
- 12 KV Cache Reduction Techniques Every AI Engineer Should Understand, Explained — blaizedsouza · 2026-09-11
- The shadow GPU capacity market is formalizing, with Meta selling excess compute to outside buyers — DavidLinthicum · 2026-09-11
- Engram's random reads don't suit SSDs; CPU-memory over NVLink could serve all 72 GPUs — bookwormengr · 2026-09-11
- 80% of the DIY LLM inference hype posters have already quit — it's brutally hard systems work — abhijithneil · 2026-09-11
- Hugging Face's Ultra Scale Playbook: a free book on training LLMs on GPU clusters — mdancho84 · 2026-09-11
- Is inference latency becoming the biggest bottleneck for production AI agents? — Euphoric_Sea632 · 2026-09-11