Xiaomi open-sources MiMo-V2.6, saying its Pro model is the top-ranked open-weight model on a major index
The omnimodal Pro and Flash models accept text, images, video and audio with million-token context. Xiaomi reports reinforcement-learning costs of $2.62 million and $850,000.
Xiaomi has released and open-sourced its MiMo-V2.6 series. The flagship MiMo-V2.6-Pro and the smaller MiMo-V2.6-Flash are natively omnimodal: they accept text, image, video and audio, and they have 1 million-token context windows. Xiaomi also released a distilled MiMo-V2.6-Distill-Qwen-9B and reinforcement-learning research resources. SiliconANGLE reports that Xiaomi also announced a Pro-UltraSpeed variant, which the company says generates output up to 20 times faster than Pro at the same quality. Citing Xiaomi's own comparison, TechNode says Pro scored 46 on the Artificial Analysis Intelligence Index. That puts it ahead of GLM-5.3 at 45 and Kimi K3 at 44, but behind leading closed models at 53. According to Xiaomi, the two models completed 30 reinforcement-learning steps in under six days using about 750,000 trajectories, at reported costs of $2.62 million for Pro and $850,000 for Flash. The release adds 3D spatial reasoning, computer-use capabilities and a desktop client. It narrows the gap between open-weight and closed models, although the benchmark comparisons come from the company.