BlockBeats News, August 21st, Max For AI revealed in a post that a leaked Benchmark screenshot of what appears to be Xiaomi's next-generation large model MiMo-V3-Pro is circulating in the community. The screenshot shows that this model focuses on the Coding Agent and General Agent scenarios, and some of the test scores are approaching top overseas models such as Claude Opus and GPT.

According to the leaked screenshot, MiMo-V3-Pro scored 72.8 points on SWE-Bench Pro, higher than GLM 5.3 with 67.9 points and Kimi K3 with 65.8 points, and close to Claude Opus 5 with 74.6 points and GPT-5.6 Sol Max with 75.4 points, a difference of less than 3 points; it scored 70.6 points on Terminal-Bench 2.0, similarly close to Claude Opus 5 with 72.0 points and GPT-5.6 Sol Max with 73.5 points. In addition, its τ3-bench score reached 76.4 points, while GPT-5.6 Sol Max scored 78.8 points.

If the above scores are eventually confirmed by the official sources and can be reproduced in the final version, MiMo-V3-Pro may enter the first echelon of global leading models. However, the authenticity of the Benchmark screenshot and the testing conditions have not yet been confirmed by Xiaomi officials, so the related data should still be considered as unconfirmed leaks. It is worth noting that Xiaomi's latest officially announced MiMo flagship models are currently MiMo-V2-Pro and MiMo-V2.5-Pro, so whether V3-Pro exists and its specific release time are still awaiting further official disclosure.