LongCat AI is an open-source AI stack from Meituan that combines large language models, coding agents, formal math, video generation, and audio-driven avatars, with branding referencing the classic Longcat internet meme. The ecosystem includes LongCat-Flash, a 560B-parameter Mixture-of-Experts (MoE) language model with about 18.6B to 31.3B parameters activated per token, pre-trained on about 20 trillion tokens in 30 days with a reported 98.48% time availability, extending to 128k tokens context and exceeding 100 tokens per second inference throughput on H800 GPUs at an output cost near 0.7 USD per million tokens. LongCat-2.0 is a 1.6T-parameter MoE model with 33B to 56B parameters activated per token, supporting a native 1 million token context and reporting a SWE-bench Pro score of 59.5. Other models include LongCat-Flash-Prover for formal mathematics producing Lean4 proofs, LongCat-Video (a 13.6B-parameter video model), and LongCat-Video-Avatar for audio-driven character animation.