Woofun AI reports that benchmark screenshots for Xiaomi's MiMo-V3-Pro have circulated online, indicating strong performance in coding and agent tasks. The leaked data shows a score of 72.8 on SWE-Bench Pro, surpassing GLM 5.3 and Kimi K3, while remaining within 3 points of Claude Opus 5 and GPT-5.6 Sol Max. Additional metrics include 70.6 on Terminal-Bench 2.0 and 76.4 on τ3-bench. Xiaomi has not verified these results or confirmed the model's existence, as current official releases remain at the V2.5-Pro level.