Meituan's LongCat 2.5 Arrives on Vercel's AI Gateway
Vercel has added Meituan's LongCat 2.5 to its AI Gateway. The model combines native visual understanding with coding and multi-step tool use, and it accepts text and image inputs with a 1M-token context window.
- Meituan's LongCat 2.5 is now available on Vercel's AI Gateway under the name meituan/longcat-2.5.
- The model supports text and image inputs with a 1M-token context window.
- Meituan highlights gains in visual programming, creative design, and 3D tasks, plus longer agent workflows from planning to execution.
- It can be used in Claude Code, Codex, Cursor, and more after running npx vercel ai-gateway setup.
- Vercel says AI Gateway reflects provider pricing with no markup and no platform fee on inference, including BYOK requests.
What LongCat 2.5 offers
According to Vercel, LongCat 2.5 is built for agents that work across complex codebases and full-project builds. It pairs native visual understanding with coding ability and multi-step tool use, so a single model can handle both what it sees and what it needs to build.
Meituan highlights improvements in visual programming, creative design, and 3D tasks. The company also points to longer agent workflows that carry engineering work from planning through execution.
The model supports text and image inputs and offers a 1M-token context window. Vercel says this is meant for code, documents, and extended task histories, which are the kinds of material long-running agent sessions tend to accumulate.
How to use it
Developers call the model through the name meituan/longcat-2.5. Vercel's example uses the AI SDK and its streamText function. The sample prompt asks the model to outline a database migration in three brief steps, and the code streams the response to standard output as it arrives. Vercel also lists Chat Completions as a supported way to call the model.
For coding agents, Vercel says users can work with Claude Code, Codex, Cursor, and more. The steps are:
- Install the latest Vercel CLI.
- Run npx vercel ai-gateway setup.
- Select meituan/longcat-2.5 in the agent.
Vercel points to its coding agents guide for details and offers a model playground where LongCat 2.5 can be tried directly. A model details page is also linked.
The AI Gateway context
AI Gateway is Vercel's unified API for calling models. According to Vercel, it also tracks usage and cost, and lets teams configure retries, failover, and performance optimizations aimed at higher-than-provider uptime. The listed features include built-in custom reporting, Zero Data Retention support, budgets for API keys, and routing rules.
On pricing, Vercel says AI Gateway reflects provider pricing with no markup. It does not charge a platform fee on inference, including on Bring Your Own Key (BYOK) requests. The source does not state LongCat 2.5's specific prices.
What it means for developers
For teams already routing models through AI Gateway, LongCat 2.5 is now another option that can be selected by changing the model name. Those who use coding agents such as Claude Code, Codex, or Cursor can point them at the model through the Vercel CLI setup. The combination of image input and a 1M-token context window makes it a candidate for tasks where screenshots, designs, or large codebases are part of the work, based on the capabilities Meituan and Vercel describe. Performance figures were not included in the announcement, so how it compares with other models on AI Gateway is not addressed.
Vercel: LongCat 2.5 now available on AI Gateway
