LongCat-2.0

LongCat-2.0 is a next-generation, trillion-parameter Mixture of Experts (MoE) large language model specifically engineered for agentic coding and complex task execution. It features a native 1-million-token context window and is the first model of its scale to be trained and deployed entirely on a domestic (Chinese) AI ASIC supercomputing cluster.

LongCat-2.0 is a next-generation, trillion-parameter Mixture of Experts (MoE) large language model specifically engineered for agentic coding and complex task execution. It features a native 1-million-token context window and is the first model of its scale to be trained and deployed entirely on a domestic (Chinese) AI ASIC supercomputing cluster.

• 011.6 Trillion parameters with dynamic MoE activation (33B–56B per token)
• 02Native 1-million-token long context window using LongCat Sparse Attention (LSA)
• 03Purpose-built for agentic coding and automated task execution
• 04First trillion-parameter model trained and inferred on domestic AI ASIC superpods
• 05135B parameter N-gram Embedding for enhanced architectural efficiency