MiMo V2.6 Pro
MiMo V2.6 Pro accepts text, images, video and audio, then returns text. It supports reasoning, structured output, tool calls and web search through Xiaomi's hosted service.
MOPD combines several kinds of teacher
The MOPD checkpoint upgrades the earlier RL release. It combines teachers trained on verifiable tasks with supervised teachers for harder-to-score domains. Prefix-based training lets the student learn a decision without recreating the entire earlier conversation.MiMo-V2.6-Pro-MOPD official model card
Inspect progress between tool calls
Practical guidance. For long tasks, review whether each action adds evidence or changes the result. Track repeated calls and completed work alongside final-answer quality. The current API and a frozen local checkpoint should be identified separately in any comparison.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API, Downloadable, Provider product |
| Model identifiers | mimo-v2.6-promimo-v2.6-pro-ultraspeedMiMo-V2.6-Pro-MOPD |
| Context window | 1,000,000 tokens Hosted Pro context limit. MiMo current model catalogue |
| Published MOPD weights | 1.02 trillion total parameters and 42 billion active parameters; MIT licence.MiMo-V2.6-Pro-MOPD official model card |
| UltraSpeed | An API service option for Pro, with access arranged through Xiaomi. It is grouped with the underlying model.MiMo current model catalogue |
| API update | The tool-call repetition fix reached the unchanged API ID on 25 September 2026 at 06:00 UTC+8.MiMo V2.6 tool-call repetition investigation |
Related models
- MiMo V2.6 Flash
Xiaomi · MiMo V2.6
A multimodal agent model that accepts text, images, video and audio and returns text.
Launch date not confirmedAPI · Downloadable · Provider product - MiMo V2.5 Pro
Xiaomi · MiMo V2.5
V2.5 Pro is a text-only mixture-of-experts release for software engineering and long agent runs.
Launch date not confirmedDownloadable - MiMo V2.6 Pro RL
Xiaomi · MiMo V2.6
A large multimodal mixture-of-experts model trained jointly on coding, agent, visual and cybersecurity tasks.
Launch date not confirmedResearch releaseDownloadable