Xiaomi has officially launched and made open-source its advanced large-scale AI model, MiMo-V2-Flash, which stands out for its remarkable power, efficiency, and blazing-fast processing speed. Boasting a staggering 309 billion parameters in total, with 15 billion actively engaged in computations, this model is meticulously crafted for Agent AI applications. In a series of rigorous and comprehensive benchmark evaluations, its performance has proven to be on par with that of DeepSeek-V3.2, a notable benchmark in the AI community.

MiMo represents Xiaomi's pioneering venture into large-scale models tailored specifically for reasoning tasks. This groundbreaking model was initially open-sourced at the end of April this year, marking a significant milestone in Xiaomi's AI journey. The substantial enhancements in its reasoning capabilities can be attributed to a series of multi-level innovations that have been meticulously integrated into its design.

Luo Fuli, the visionary leader behind the MiMo project, is set to make her much-anticipated public debut at tomorrow's Human-Car-Home Ecosystem Partner Conference. Her presence is expected to shed further light on the future trajectory and potential applications of this cutting-edge AI model.