LongCat 2.0 from Meituan now on OpenRouter (1,049k context, $0.30/M in)
TL;DR
Meituan released LongCat 2.0, a sparse mixture-of-experts model with 48B active parameters out of 1.6T total, on OpenRouter.
What changed
LongCat 2.0 from Meituan launched on OpenRouter with 1,049k context length at $0.30 per million input tokens and $1.20 per million output tokens. The sparse mixture-of-experts model activates 48B parameters from a 1.6T total parameter count. Developers and Vibe Builders now access it directly through the platform for coding and agentic workflows.
Why it matters
Basic Users gain an affordable option for long-horizon tasks while Developers handle repository-level changes at scale with the expanded context window. Vibe Builders can test agentic setups against models like Claude 3.5 Sonnet which charges higher rates on similar coding benchmarks. This pricing and context combination supports practical experiments in large codebases without immediate budget spikes.
What to watch for
Compare outputs against Gemini 1.5 Pro on the same OpenRouter endpoint for consistency in multi-file edits. Developers should run a verification test by submitting a 500k token repository prompt and checking token usage logs in the dashboard.
Who this matters for
- Vibe Builders: Test agentic workflows on OpenRouter to compare this model against Claude 3.5 Sonnet pricing.
- Developers: Run a 500k token repository prompt to verify multi-file edit consistency and check token logs.
Harsh’s take
Meituan entering the long-context arena with a 1.6T parameter MoE model at these price points is a direct shot at established players. At thirty cents per million input tokens, running massive codebases through an LLM is no longer a luxury reserved for well-funded teams. Do not just trust the benchmark claims blindly.
Developers need to actively test how this sparse MoE handles needle-in-a-haystack retrieval at the 1M token limit compared to Gemini. If the retrieval holds up, this becomes an instant default for budget-conscious agentic pipelines.
by Harsh Desai
About Ito
View the full Ito page →All Ito updatesGo deeper
More from Ito
- PricingGPT-5.6 Sol nearly matches Fable 5 on aggregated benchmarks at one-third the cost
OpenAI's GPT-5.6 Sol scores 59 points on the Artificial Analysis Intelligence Index, one point behind Claude Fable 5. At $1.04 per task it costs one third as much as Anthropic's top model and leads in agentic coding.