← Library · Frontier

Meituan Releases LongCat-2.0, a 1.6 Trillion Parameter Coding Model

Meituan has open-sourced LongCat-2.0, a 1.6-trillion-parameter Mixture-of-Experts model designed for agentic coding. It activates approximately 48 billion parameters per token and boasts a 1-million-token context window. Remarkably, LongCat-2.0 was trained and served entirely on Chinese-made Ascend chips, without any Nvidia hardware.

Why it matters

This release provides a near-frontier agentic coding model under an MIT license, significantly reducing per-token costs and removing vendor lock-in for developers. Its large context window allows reasoning over entire code repositories, making advanced agentic workflows more feasible and cost-effective.

Learn one new AI thing every day.

Daily Deck sends you seven plain-English cards like this every morning. Free.

Start free