The most recent assessment results of the Chinese large model benchmark, SuperCLUE, for March 2026, have been disclosed, featuring participation from 22 mainstream models from both domestic and international sources. In the overall score rankings, overseas closed-source models demonstrated outstanding performance, with offerings from Anthropic, Google, and OpenAI securing the top three positions. ByteDance's Doubao model emerged as the domestic leader, achieving a score of 71.53 and breaking into the global elite. It trailed GPT-5.4 by a mere 0.95 points in the total score and ranked among the top five globally in the agent task planning category. Two models from Xiaomi also made it onto the list, with MiMo-V2-Pro distinguishing itself among closed-source models and MiMo-V2-Flash showcasing promise in areas such as code generation. Domestic models, as a whole, delivered strong performances, particularly excelling in the open-source arena, where they claimed the top three positions on the open-source leaderboard.