Available on request
MiMo API:
text, image, video and audio in one model.
Xiaomi’s MiMo-V2.6-Flash takes all four input types in a single model, with a 1M-token context and an MIT licence. If you would rather not stitch a speech model, a vision model and a language model together, tell us your volume and we’ll talk through adding MiMo to the Xinference Model API.
One model instead of three
What “omnimodal” covers
Images and video
A dedicated vision encoder of about 681M parameters reads screenshots, charts, photos and video frames.
Audio
An audio tokenizer and patch encoder take speech and sound directly, without a separate transcription step first.
Long agent runs
Xiaomi trained it with one mixed reinforcement-learning run across coding, general agents, visual tasks and cybersecurity, and gave it a 1M-token context for long repositories and tool traces.
Model facts
MiMo-V2.6-Flash at a glance
| Maker | Xiaomi (MiMo team) |
| Architecture | Sparse Mixture-of-Experts: 309B total, 15B active per token |
| Context | 1M tokens Model maker’s figure |
| Input | Text, image, video and audio |
| Licence | MIT |
Source: Xiaomi’s MiMo-V2.6-Flash model card, checked 23 October 2026. Xiaomi MiMo and the Xiaomi MiMo logo are trademarks of Xiaomi, shown only to identify the model. Their use does not imply endorsement.
MiMo API questions
Is MiMo free for commercial use?
MiMo-V2.6-Flash is released under the MIT licence, one of the most permissive licences available. It allows commercial use, modification and redistribution, as long as the copyright notice is kept.
Can MiMo understand audio?
Yes. Xiaomi lists text, image, video and audio as input types for MiMo-V2.6-Flash, with a dedicated audio encoder alongside the vision encoder.
How popular is MiMo?
On OpenRouter’s public rankings for the 30 days to 22 October 2026, MiMo-V2.6-Flash was the fourth most-used model by tokens processed.
Is MiMo on the Xinference Model API?
On request. The Model API’s standard models, with public per-token pricing, are Kimi K3 and DeepSeek V4.1 Flash. For MiMo, tell us your expected volume.
Last updated