Updated

MiMo-V2.6-Pro and MiMo-V2.6-Flash

Xiaomi open-sourced the MiMo-V2.6-Pro and MiMo-V2.6-Flash multimodal LLMs, with Pro scoring 46 on the AI Composite Intelligence Index to become the top open-weight model globally. Both models use scaled reinforcement learning for recursive self-improvement and are available as open weights.

By No Code MBA · Facts checked

Provider
Xiaomi
Announcement date
Status at last check
Generally available
Model family
MiMo-V2.6
Preceding model
MiMo-V2.5-Pro
Capabilities
Text, Image, Coding

What changed

MiMo-V2.6-Pro and MiMo-V2.6-Flash are Xiaomi's natively multimodal successors to the MiMo-V2.5 series, trained with scaled reinforcement learning that supports 1M token context during training. Xiaomi reports Pro scores 46 on the Artificial Analysis Intelligence Index, the top open model by its count, and matches Claude Opus 5 and GPT-5.6 Sol on most agent benchmarks. On DeepSWE v1.1, Xiaomi reports Flash rose from 48.8 to 65.7 and Pro from 58.4 to 72.6.

Access and availability

Open weights and a technical report are in the XiaomiMiMo Hugging Face collection. The Xiaomi MiMo API serves mimo-v2.6-pro, mimo-v2.6-flash and mimo-v2.6-pro-ultraspeed (model IDs must be lowercase).

Official access documentation

What you can build

Try MiMo-V2.6-Pro or MiMo-V2.6-Flash through their open-weight releases for cost-efficient multimodal tasks like image understanding, 3D perception, and computer use, especially if you want to avoid per-token API costs from closed-source providers.

Limitations

Xiaomi acknowledges a gap with the strongest closed models, naming Claude Fable 5.1 and GPT-6 Astra. Parameter counts, inference context length and license were not stated in the announcement, and V2.5-Pro and V2.5 model IDs stopped working on October 21, 2026.

Sources and release history