Chinese tech giant Meituan has open-sourced LongCat-2.0, a massive artificial intelligence model with 1.6 trillion parameters and a one million token context window, which the company claims is the first of its scale trained entirely on domestic Chinese chips. Meituan described it as "the industry's first trillion-parameter model to complete end-to-end training and inference on a 50,000-chip domestic compute cluster." LongCat-2.0 is available under an MIT license on GitHub and Hugging Face. The model was the computational engine behind "Owl Alpha," which topped developer charts on OpenRouter for two months before Meituan revealed its identity. Meituan claims LongCat-2.0's performance matches Google's Gemini 3.1 Pro, released in February, and scored 59.5 on SWE-bench Pro, surpassing reported scores for GPT-5.5. During its unbranded period on OpenRouter, it processed approximately 10.1 trillion monthly tokens. The system uses a Mixture-of-Experts architecture with an average of 48 billion parameters active per token. Commercial API pricing is set at $0.75 per million input tokens and $2.95 per million output tokens, with promotional discounts reducing these to $0.30 and $1.20 respectively. The announcement comes as China accelerates efforts to develop domestic chip capabilities in response to US export controls on advanced semiconductors.