LongCat 2.5 by Meituan added to BenchGecko
Meituan's 1.6T parameter multimodal agent model is now tracked on BenchGecko · $0.30 input and $1.20 output per million tokens · 1M token context.
What was added
LongCat 2.5 by Meituan is now listed on BenchGecko. Meituan released it on November 2, 2026 as a natively multimodal model built for long horizon agent work and coding. It succeeds LongCat 2.0, which is already tracked.
Size and context
Meituan describes LongCat 2.5 as a mixture of experts model with 1.6 trillion total parameters and about 48 billion activated per token. The API accepts a context window of 1,048,576 tokens, and a single response can run up to 262,144 tokens. Vision input is handled inside the same token stream as text rather than through a separate pipeline.
Pricing
The LongCat API lists pay as you go pricing of $0.30 per million uncached input tokens, $0.006 per million cached input tokens and $1.20 per million output tokens. Meituan labels these as limited time launch prices, so BenchGecko will update the listing when they change. At these rates, a task with 1 million input tokens and 500,000 output tokens costs about $0.90 before any cache discount.
Availability
LongCat 2.5 is available today in the LongCat chat app and through an API that accepts both OpenAI and Anthropic request formats. Meituan says weights for GitHub and Hugging Face will follow, so BenchGecko lists it as a proprietary API model until they are published.
Benchmarks
Meituan reports that LongCat 2.5 scores between Claude Opus 4.8 and Claude Opus 5 on most mainstream benchmarks and ahead of Claude Opus 5 on some. These are vendor results. BenchGecko will add independent scores to the model page as they arrive.
Source: BenchGecko
Written automatically from BenchGecko's tracked data for this period; every number comes from the changelog, model, mindshare and pricing datasets. Checked against the source data before publishing.