AI BriefingMeituan Launches LongCat-2.0 API with 1.6T Parameters
Key point
Meituan has released LongCat-2.0, a 1.6-trillion-parameter MoE model, in API form, strengthening agentic coding and long-context processing capabilities.
Details
Meituan has unveiled LongCat-2.0, an evolution of the existing LongCat-2.0-Preview. This model is designed with a MoE (Mixture-of-Experts) architecture totaling 1.6 trillion (1.6T) parameters, with approximately 48 billion parameters activated per token.
LongCat-2.0 offers the following key features:
- A 1 million token context window
- Support for output lengths of up to 128K tokens
- Efficient long-context processing via LongCat Sparse Attention (LSA) technology
- Optimization for agentic coding, tool use, automated workflows, and complex instruction following
This model was pretrained on over 35 trillion tokens and was trained and deployed on an AI ASIC superpod. Developers can use this model in various workflows such as Claude Code and OpenClaw through OpenAI- and Anthropic-compatible APIs.
Currently, the model weights have not been released, making this a hybrid launch, available only for paid use via the API. Pricing is set at $0.75 per 1 million input tokens (without cache) and $2.95 per 1 million output tokens.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Step 5 Preview Released: 600B Parameter MoE Model Highlights Agentic Tasks and Financial Analysis Performance