Xiaomi Open-Sources MiMo-V2.6 Series, Boosting Research Efficiency 10x

Chinese tech company Xiaomi has officially released and open-sourced the Xiaomi MiMo-V2.6 series, which includes two native omni-modal models: Pro and Flash. Xiaomi calls this a key step in exploring the RSI (recursive self-improvement) path.

MiMo-V2.6-Pro has already taken on the role of "Co-Scientist" in real research tasks. In new materials research, the model helped researchers design metal-organic framework materials by searching literature, proposing candidates, and running simulation tools to analyze and screen material performance. It cut a research cycle from about one month down to 2-3 days, a 10x efficiency boost. Dou Jinhu, a researcher at Peking University, said it reached the level of a trained PhD researcher.

In mathematical formal proof, MiMo-V2.6-Pro helped researchers complete the full formal proof of the main theorem from the classic Li-Yorke paper in Lean 4, with over 6,000 lines of source code passing kernel verification.

The MiMo-V2.6 series keeps the same API pricing as V2.5: Flash costs 1 yuan per million input tokens and 2 yuan per million output tokens; Pro costs 3 yuan and 6 yuan, with a 99% cache discount. At the same intelligence level, the price is only 1/20 to 1/60 of overseas models.