Xiaomi releases open-weight MiMo-V2.6 models with 1-million-token context

Xiaomi released MiMo-V2.6-Pro and MiMo-V2.6-Flash with MIT-licensed weights, API access and support for text, image, video and audio inputs. The flagship Pro model has 1.02 trillion total parameters, activates 42 billion per token and supports a 1-million-token context window. Artificial Analysis scored Pro at 46 on its Intelligence Index, the highest result among open-weight models on that leaderboard. Xiaomi also published a technical report, reinforcement-learning environments and training code. Its own tables show mixed performance: Pro scored 71.9 on DeepSWE v1.1, below DeepSeek V4.1 Flash, Claude Opus 5 and GPT 6 Astra, while API prices remain unchanged from MiMo-V2.5.