Meituan’s LongCat-2.5 pairs multimodal input with long context

Meituan released LongCat-2.5 on November 1, 2026, bringing native multimodal capabilities and a 1.0M-token context window to its language model lineup. The model supports text, image and video inputs and produces text output, with a focus on long-horizon agentic tasks.

The Mixture-of-Experts architecture has about 1.6T total parameters, with roughly 48B activated per token. It builds on LongCat-2.0 with stronger native visual understanding, combining support for visual inputs with the capacity to handle lengthy prompts and conversations.

Access through Meituan starts at $0.300 per million input tokens and $1.20 per million output tokens. Reused prompt prefixes cost $0.0060 per million cached input tokens, providing discounted processing when prompt tokens are served from the provider’s cache. Official documentation and a playground are available through longcat.chat, while a technical report is available through longcat.ai.