Meituan Open-Sources LongCat-2.0, Releases Domestic Chip Inference Code
Original title:正式开源!美团 LongCat-2.0 同步开放国产卡推理代码
AI summary
Meituan's trillion-parameter large model LongCat-2.0 is now officially open-source, with 1.6T total parameters and about 48B average activated parameters, designed for real Agentic Coding tasks. The model introduces the LongCat sparse attention mechanism and N-gram Embedding to improve long-context processing efficiency and token-level representation. LongCat-2.0 is the industry's first trillion-parameter model to complete inference on a 50,000-GPU domestic compute cluster, with deep co-optimization from model architecture and chip adaptation to deployment strategy, supporting efficient million-token context inference. The open-source release also provides BF16, FP8, INT8 and other precision versions along with inference code optimized for domestic compute.
Key Takeaways
- Meituan's trillion-parameter large model LongCat-2.
- 0 is now officially open-source,with 1.
- 6T total parameters and about 48B average activated parameters,designed for real Agentic Coding…
- The model introduces the LongCat sparse attention mechanism and N-gram Embedding to improve…
- LongCat-2.
Quick Facts
- Published
- Updated
- Source
- 公众号:龙猫LongCat(美团)
- Category
- Model Releases
- AI score
- 74
国产算力跑万亿模型的第一份完整开源答卷,代码、芯片适配细节全公开,做国产大模型落地的直接抄作业。
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 Open-Sources LongCat-2.0, Releases Domestic Chip Inference Code. July 9, 2026. https://new.nat6.net/en/news/cmrees13p006wihgvrokrpgpy/