Meituan has added LongCat-2.5 Preview to the LongCat API platform. The model is available through the API in OpenAI- and Anthropic-compatible formats and is intended, according to the company’s description, for programming and long-running agentic workflows. The Preview status means that the parameters and availability may change.
Confirmed parameters
The API documentation for LongCat-2.5 Preview specifies a context window of up to 1M tokens and a maximum response length of 131 072 tokens. Requests can be sent via an OpenAI-compatible endpoint or the Anthropic Messages interface; reasoning mode is enabled with a request parameter. A LongCat API key is required for the first request.
Meituan describes the model as multimodal and designed for long-running tasks involving tools. This is the developer’s statement about the model’s intended use. The documentation contains no public independent benchmark that would allow LongCat-2.5 Preview’s quality to be compared with other agentic models.
Pricing depends on the access point
The LongCat-2.5 Preview listing on Vercel AI Gateway displayed an introductory rate of $0.30 per million input tokens and $1.20 per million output tokens; a separate price for cached reads was also listed there. The listing marks these prices as discounted. This is the rate charged by a specific intermediary, not a guarantee of the price for direct access through Meituan’s platform. Before integrating, check the current amount and billing terms in the service through which your requests will be routed.
What to check before going live
- Check the current context and response limits in the documentation: they may change for a Preview model.
- Check which input types are supported by the selected API method. The description “multimodal” by itself does not replace verifying the request format.
- Test the model on your own typical tasks and measure the cost of the agent’s full cycle, including repeated calls.
- Do not move a critical process to a Preview model without a backup provider and saved test results.
The parameters and request examples are published in the LongCat API documentation; the gateway’s current price can be checked in the Vercel model listing.
Compare models before you start
The service sets its plans, limits and model catalog. If they differ from this article, contact us so we can update it and record a new review date.
Browse modelsAffiliate link: your price stays the same and the project earns a commission.
