Meituan’s LongCat 2.0 — a 1.6-trillion-parameter open coder trained on Chinese chips — lands on OpenRouter

On July 20, 2026, OpenRouter officially listed LongCat 2.0, Meituan’s MIT-licensed mixture-of-experts model (1.6T total / ~48B active parameters, 1M-token context) that had been serving in stealth as “Owl Alpha”; the weights were open-sourced July 6. Meituan says it was trained and served entirely on a 50,000-card domestic Chinese ASIC cluster — a vendor claim — and routed pricing is $0.30 per million input tokens and $1.20 per million output.

What it means A near-frontier, MIT-licensed agentic coder at roughly a tenth of frontier pricing — and a supply-chain signal that trillion-parameter training no longer requires NVIDIA hardware.