Xiaomi open-sources MiMo-V2.6 under MIT, with its RL training code
Xiaomi released MiMo-V2.6-Pro and MiMo-V2.6-Flash as open weights under the MIT licence on September 22, 2026. Pro is a mixture-of-experts model with 1.02 trillion parameters and 42 billion active per token; both take text, images, video and audio with a 1M context window. Xiaomi also published more than 7,000 reinforcement learning environments and the training code. Through OpenRouter, Pro costs $0.435 per million input tokens and $0.87 output.
Why it matters
At 46 on the Intelligence Index it is the highest-scoring open-weight model, level with Grok 4.7. Publishing the RL environments means the method can be checked and reused, which is rare at this size.
Who should care
Teams that need to self-host, and anyone doing their own post-training
What you can do
Try Pro through a router before planning hardware: serving 1.02T parameters needs a multi-GPU setup.