XiaomiMiMo V2.6
MiMo V2.6 Distill Qwen 9B
This 9B model is a supervised fine-tune of Qwen3.5-9B on MiMo-generated data. Xiaomi released it as a starting point for research into reinforcement learning for agents.
Research release
Separate distillation from later reinforcement learning
Practical guidance. Use the published SFT checkpoint as the baseline when studying an additional training method. Report the base model, teacher data and new training stage separately. That makes improvements easier to attribute and reproduce.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | MiMo-V2.6-Distill-Qwen-9B |
| Ancestry | Qwen3.5-9B; it is not a small MiMo architecture trained from scratch.MiMo-V2.6-Distill-Qwen-9B official model card |
| Training scope | Coding, general agents, visual coding and cybersecurity.MiMo-V2.6-Distill-Qwen-9B official model card |
| Licence | MIT in the Xiaomi release.MiMo-V2.6-Distill-Qwen-9B official model card |
Related models
- MiMo V2.6 Flash RL
Xiaomi · MiMo V2.6
A smaller MiMo V2.6 mixture-of-experts model trained for multimodal understanding and long-context agent work.
Launch date not confirmedResearch releaseDownloadable - 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