Xiaomi

AI · Open-Source Models · Large Language Models

Launch

Xiaomi's MiMo-V2.6-Pro launches as top open-weight AI model, ties Grok 4.7

The model is free to download and fine-tune under an MIT license, undercutting rival API pricing by roughly fivefold.

• Xiaomi released and open-sourced MiMo-V2.6-Pro and MiMo-V2.6-Flash, with Pro scoring 46.32 on Artificial Analysis' Intelligence Index, tying xAI's Grok 4.7 and beating Grok 4.6 and DeepSeek V4.1 Pro.
• Xiaomi, better known for smartphones and EVs than frontier AI, built the model through its MiMo research team, led by Luo Fuli, a former DeepSeek researcher.
• MiMo-V2.6-Pro is a sparse mixture-of-experts model with 1.02 trillion total parameters but only 42 billion active per token, natively handling text, image, audio and video with a 1-million-token context window.
• Xiaomi charges $0.435/$0.87 per million input/output tokens for Pro, roughly a fifth of Grok 4.7's price, while spending about $2.62 million and under six days on the model's final reinforcement-learning stage.
• The jump is steep: Xiaomi's prior flagship, MiMo-V2.5-Pro, scored just 26 on the same Intelligence Index months earlier, nearly doubling to 46.32 with V2.6-Pro.
• A hardware and consumer-electronics company topping open-weight AI leaderboards at a fraction of proprietary API costs signals that the intelligence gap between Chinese open models and Western closed frontier labs is narrowing fast.