Meituan LongCat Flash

The LongCat Flash family introduces large-scale Mixture-of-Experts language models with dynamic computation, activating a subset of parameters based on contextual demands. It emphasizes efficient training, strong reasoning, and advanced agentic capabilities through scalable reinforcement learning and domain-parallel optimization.

The LongCat Flash family introduces large-scale Mixture-of-Experts language models with dynamic computation, activating a subset of parameters based on contextual demands. It emphasizes efficient training, strong reasoning, and advanced agentic capabilities through scalable reinforcement learning and domain-parallel optimization.