← All models
Meituan LongCat 2.5
1 providersLongCat-2.5 is Meituan's newly released (Nov 2, 2026) high-performance agentic, natively multimodal LLM with 1.6T total parameters (MoE, ~48B activated per token), trained entirely on domestic compute. It supports a 1M-token context window, up to 128K output tokens, native chain-of-thought reasoning, function calling and vision understanding, with both OpenAI- and Anthropic-compatible APIs. Benchmarks place it alongside top-tier international closed models — ideal for agents, deep reasoning, long documents, coding and visual understanding.
NewLimited-time PriceReasoningAgentLong ContextCodeMeituan
Meituan's latest 1.6T MoE natively multimodal agentic model: 1M context, native reasoning, tool calling and vision.
Compare providers
Same underlying model, served by different providers. Compare pricing, context and latency, then call via the exact model_id shown.
| Provider | Input / M | Output / M | Context | Latency | Status | model_id | Detail |
|---|---|---|---|---|---|---|---|
| ¥2.0 | ¥8.0 | 1000K | 2586ms | Available | LongCat-2.5 | View → |