[Meituan Releases Trillion-Parameter Model LongCat-2.0, the First Trillion-Parameter Model Fully Trained on a Domestic Computing Cluster] According to Deep Tide TechFlow, on June 30, Meituan officially announced the launch of its next-generation large model, LongCat-2.0, along with its open-source release. The model boasts a total of 1.6 trillion parameters and is the industry's first trillion-parameter model to complete full-process training and inference on a domestic computing cluster with 50,000 GPUs. It natively supports 1M ultra-long context and is primarily focused on code understanding, generation, and execution in Agentic Coding scenarios. On the technical side, LongCat-2.0 employs the LongCat Sparse Attention (LSA) mechanism, reducing the computational complexity of long texts from quadratic to linear. It achieves token-level dynamic activation (33B~56B) through a zero-computation expert mechanism and integrates the MOPD architecture to combine the expertise of Agent, Reasoning, and Interaction. In terms of training efficiency, the team spent three years overcoming challenges in adapting to domestic computing power, reducing the average daily failure rate by over 70% per month, increasing training MFU by 1.5 times, and achieving a steady-state daily throughput of over 1 trillion tokens/day. For performance evaluation, LongCat-2.0 scored 59.5 on SWE-bench Pro, surpassing Gemini 3.1 Pro (54.2), GPT-5.5 (58.6), and Claude Opus 4.6 (57.3). It also achieved a score of 79.9 on BrowseComp, reaching the level of cutting-edge closed-source models.