【#Tech24H】On June 30, Meituan officially released its next-generation trillion-parameter large model, LongCat-2.0, and open-sourced it to the public. This trillion-parameter model has completed end-to-end training and inference on a 50,000-card domestic GPU cluster. It has a total of 1.6T parameters, with an average activated parameter of about 48B and a dynamic range from 33B to 56B. The model was pretrained on more than 30T tokens, including Chinese, English, multilingual text and code corpus, and natively supports a 1M-token ultra-long context window.