Meituan releases a 1.6-trillion-parameter agent model that never touched an Nvidia GPU - The Decoder

The LongCat team at Chinese food delivery company Meituan has released the weights for LongCat-2.5, one week after the model's official unveiling. LongCat-2.5 is built for agents that work through long, multi-step tasks autonomously, particularly coding. According to Meituan, it comes close to top Western models on several benchmarks.

LongCat-2.5 is a mixture-of-experts model that activates only a subset of its specialized subnetworks for each token it processes. Out of 1.6 trillion total parameters, roughly 48 billion are active per token. Through the API, the model handles up to one million tokens at once and generates responses up to 128,000 tokens long.

Multimodal capabilities

LongCat-2.5 builds on its predecessor LongCat-2.0 and adds image understanding. Unlike many multimodal models that bolt on a separate image encoder, LongCat-2.5 converts images into tokens that the model processes alongside text. The approach draws on the research model LongCat-Next and aims to improve visual coding, design, and 3D tasks.

LongCat-2.5 breaks images down at their native resolution into discrete tokens that the model processes alongside text. For reinforcement learning, the team used real software tasks from GitHub pull requests along with recreated agent failures. To ensure the model actually solved tasks during training, developers blocked internet access, removed Git history and cached files, and had a language model check solutions for shortcuts.

Chinese hardware optimization

Meituan quantizes the experts' weights and activations to 4-bit precision, which Chinese chips can process natively. In load tests, throughput rises 43.3 percent for prompt processing and 42.77 percent for response generation compared with an 8-bit version, while accuracy stays on par with the uncompressed model.

Performance benchmarks

The LongCat team reports 88.4 points on Terminal Bench v2.1, a benchmark that measures how well agents solve command-line tasks. Meituan lists Claude Opus 5 at 89.1, GPT-5.6 Sol at 88.8, and Deepseek-V4.1-Flash leading at 90.6. On DeepSWE tests for bug fixing in code repositories, LongCat-2.5 scores 64.6 points, behind most other models except Claude Opus 4.8.

Terminal Bench v4.0, the harder variant, puts LongCat-2.5 at 38.5 points, slightly above GPT-5.6 Sol's 37.3 but well below Opus 5's 53.9. The model posts 91.9 on the web search benchmark BrowseComp, the highest score in the table, though Meituan ran this test in its own environment.

Open-source availability

The weights are available under the MIT license, allowing commercial use, modification, and redistribution. Users can also try LongCat-2.5 in a chat interface at https://longcat.ai/. Its API supports both the OpenAI and Anthropic formats and plugs into agent harnesses like Claude Code, Codex, and Hermes. At launch, Meituan is charging a limited-time promotional price equivalent to roughly €0.27 per million input tokens and €1.07 per million output tokens.

Meituan says it trained the predecessor LongCat-2.0 in late June entirely on a cluster of more than 50,000 Chinese AI chips. Chinese open-weight models like Kimi K3.1 and GLM-5.3 have largely caught up with top US models on common benchmarks. Open-weight models optimized for Chinese chips from the start make that hardware more attractive outside China too.