Updated

LongCat-2.5 Preview

Meituan launched LongCat-2.5 Preview on its API platform, adding native image understanding to its long-context model. It keeps a 1.6T parameter MoE architecture with about 48B active parameters, a 1M token context window, and 128K token output, and targets long-horizon agent tasks across browsers, terminals, and desktop apps.

By No Code MBA · Facts checked

Provider
Meituan
Announcement date
Status at last check
Limited access or preview
Model family
LongCat
Preceding model
LongCat-2.0
Capabilities
Text, Image, Coding

What changed

LongCat-2.5 Preview adds native image understanding to Meituan's LongCat line, supporting cross-modal Q&A, summarization and visual reasoning. Meituan's API docs list a 1M token context window and 128K token maximum output. Meituan's changelog cites stronger code generation and automated programming, and compatibility with Claude Code, Hermes, OpenClaw, OpenCode and Kilo Code.

Access and availability

Available on the LongCat API Platform as LongCat-2.5-Preview, with OpenAI-format (https://api.longcat.chat/openai) and Anthropic-format (https://api.longcat.chat/anthropic) endpoints. Weights have not been released.

Official access documentation

What you can build

Test LongCat-2.5 Preview through Meituan's API for long-context agent work such as reading and modifying a large codebase, since its endpoints are compatible with Anthropic and OpenAI style calls.

Limitations

It is labeled a preview, and Meituan has published no benchmark scores for it. Unlike LongCat-2.0, it is API-only.

Sources and release history