China's Meituan released the open-source coding model LongCat-2.0 on June 30, 2026. LongCat-2.0 is a 1.6-trillion-parameter Mixture-of-Experts (MoE) model trained from scratch on a 50,000-chip cluster of domestic processors, with no U.S. accelerators. It is specialized for agentic coding, supports a 1-million-token context window, and is published under the meituan-longcat organization on Hugging Face. ASAP summarizes the facts in a direct-answer format based on primary reporting from Crypto Briefing and Reuters.

Why It Was Trained From Scratch on 50,000 Domestic Chips

Meituan said on June 30, 2026, that LongCat-2.0 was trained from scratch on a cluster of 50,000 domestic chips. According to Crypto Briefing, the training used no U.S. export-restricted accelerators such as Nvidia's A100 and H100 or AMD's MI300X. With the United States tightening advanced-chip exports to China, training a 1.6-trillion-parameter model on domestic hardware alone is the central signal of this release.

An MoE Architecture Specialized for Agentic Coding

LongCat-2.0 is a 1.6-trillion-parameter Mixture-of-Experts model that dynamically activates 33 to 56 billion parameters per token. It is designed to handle real-world coding tasks more efficiently and reliably, and it accepts inputs of up to 1 million tokens to work with ultra-long documents and large codebases. By activating only a fraction of the full 1.6 trillion parameters, the model keeps inference cost low relative to its size.

Item	Value
Release date	June 30, 2026
Total parameters	1.6 trillion (MoE)
Active per token	33–56 billion
Context window	1 million tokens

Benchmark Scores and Meituan's Claims

Meituan reported that LongCat-2.0 scored 59.5 on the SWE-bench Pro coding benchmark and 70.8 on Terminal-Bench. Meituan claimed the model matched or exceeded leading proprietary models such as Google's Gemini, OpenAI's GPT-5.5, and Anthropic's Claude Opus on some coding and agent benchmarks. The benchmark figures are Meituan's own reported numbers, and independent verification was not yet available at the time of release.

Open-Source Release in the Context of U.S. Export Controls

Meituan made LongCat-2.0 downloadable as open source under the meituan-longcat organization on Hugging Face. Releasing a large model trained on 50,000 domestic chips for free, while the United States restricts advanced AI-chip exports to China, illustrates the self-reliance strategy of China's AI camp. Earlier, players such as DeepSeek similarly used open-source releases to broaden their developer base amid the export-control environment.

Summary

Meituan released LongCat-2.0, a 1.6-trillion-parameter open-source coding model trained on 50,000 domestic chips, on June 30, 2026. The core point is that a large MoE model trained without U.S. accelerators was released for free for agentic coding, while its claimed benchmark advantage remains subject to independent verification.

---

Sources: Crypto Briefing · AOL/Reuters · Hugging Face (reported June 30, 2026)