[Meituan Releases LongCat-2.0, Test Version Ranks Among Top Three in Global Total Calls on OpenRouter] According to Golden Finance, on June 30, Meituan released its new generation foundational large model, LongCat-2.0. This is the industry's first trillion-parameter large model to complete the entire training and inference process using domestic computing power, with a total parameter scale of 1.6 trillion. Previously, the test version of this large model had already ranked among the top three globally in total calls on OpenRouter. Specifically, it ranked first globally in monthly calls on Hermes and second globally in monthly calls on Claude Code, second only to Claude Opus 4.8. Community feedback indicates that the LongCat-2.0 test version is close to Claude Opus 4.6 but lags behind the latest Claude Opus 4.8. LongCat-2.0 represents the largest training task completed using domestic computing power to date, with a peak scale exceeding 50,000 GPUs. Its implementation demonstrates that domestic computing power is now capable of supporting the full-process training of cutting-edge models. The LongCat team has officially announced plans to open-source core technologies such as the Infra framework, inference engine, and model parameters in the near future to give back to the global developer community.