Xiaomi open-sources MiMo-V2.6 models after scaling reinforcement learning

以下正文同步自 动点科技 TechNode，版权归原站所有，已转换为易读排版。

Xiaomi has released and open-sourced the MiMo-V2.6 series, including the native multimodal MiMo-V2.6-Pro and MiMo-V2.6-Flash models. Xiaomi also released MiMo-V2.6-Distill-Qwen-9B and research resources for reinforcement learning.

MiMo-V2.6-Pro scored 46 on the Artificial Analysis Intelligence Index, ahead of Kimi K3 at 44 and GLM-5.3 at 45, according to Xiaomi’s cited comparison. The company described it as the highest-ranked open-weight model on the index, while noting that leading closed-source models scored 53.

The two models completed 30 reinforcement-learning steps in less than six days, using about 750,000 trajectories at reported costs of $2.62 million for Pro and $850,000 for Flash. The release also adds 3D spatial reasoning, multimodal perception, computer-use capabilities and a desktop client. [Xiaomi MiMo, in Chinese]

正文由 FLUX 从来源站点 RSS 同步，内容未经改写；遇到排版缺失或需要图片、视频时请以原文为准。