Meituan has launched LongCat-2.5-Preview, expanding its LongCat model line with native multimodal capabilities and a focus on long-horizon agent workloads. The preview keeps the same core scale disclosed for LongCat-2.0, including 1.6T total parameters, about 48B activated parameters, and a 1M-token context window, while extending execution to longer task chains across terminals, browsers, desktop software, spreadsheets, and design tools.
The model is now available through Meituan’s official API and is compatible with both OpenAI and Anthropic interfaces. Meituan said the maximum output length is 128K tokens. It also published integration methods for Codex, OpenCode, OpenClaw, and CatPaw, which means existing agent tools can call the model directly.
For rollout, Meituan is granting 5 million free tokens to existing users so they can test the new model, while current token packages remain valid. The company has not yet released benchmark data for LongCat-2.5, and there is still no public comparison showing how much GUI operations and long-process task performance have improved.
Meituan has launched LongCat-2.5-Preview, adding native multimodal capabilities and aiming the model at long-horizon agent tasks.
The preview keeps the same figures previously disclosed for LongCat-2.0: 1.6T total parameters, about 48B activated parameters, and a 1M-token context window. On top of that, the new version is designed to carry out longer task sequences across terminals, browsers, desktop software, spreadsheets, and design tools.
API access and compatibility
LongCat-2.5-Preview has been connected to Meituan’s official API. It is also compatible with OpenAI and Anthropic interfaces, with a maximum output length of 128K tokens.
Meituan also released connection methods for Codex, OpenCode, OpenClaw, and CatPaw, allowing existing agent tools to invoke the model directly.
User access and missing benchmarks
As part of the rollout, Meituan is giving all existing users 5 million free tokens to try the new model. Existing token packages will continue to work.
The company has not published benchmark results for LongCat-2.5 so far. There is still no public data showing how much GUI operations and long-process task execution have improved.
This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan. Disclaimer:
The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.
Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.
