Text and chat · Xiaomi
Xiaomi MiMo
Xiaomi's open models that understand text, images, audio and video, with 1M context, in private mode and very cheap.
Private Low costWhat sets it apart
Xiaomi’s MiMo models are open, MIT-licensed mixture-of-experts models with a 1M token context that accept text, images, audio and video. The Flash variant keeps 15B active parameters, so it is very cheap while scoring high among open models.
- MIT license, open weights, 1M context
- Reads text, images, audio and video
- Flash: 309B total, only 15B active
Strengths and limits
Where it shines
- Reads images, audio and video at a very low price
- 1M token context, tool calling, private mode
Think twice when
- Less widely tested than the bigger families
- 1M token context
- Input: text, image, audio, video
- Open weights (MIT license)
Use cases
Summarize a recorded video call
Watch this recording of our product review and list the decisions made, who owns each one, and the deadlines mentioned.
V2.5 accepts audio and video, and the context holds long recordings.
Cheap agent runs over many documents
Go through these 150 supplier contracts and flag any with an automatic renewal clause longer than 12 months.
Flash combines a 1M context with a very low per-token price.
Against its rivals
- Qwen open modelsQwen open models are the nearest open rivals, with similar multimodal reach.#23 of 90 on Artificial Analysis Intelligence Index (40 points)
- DeepSeekDeepSeek competes on price, but mostly without audio or video input.#24 of 90 on Artificial Analysis Intelligence Index (39 points)
- KimiKimi suits long agent runs and costs more per token.#15 of 90 on Artificial Analysis Intelligence Index (44 points)
Version history
Each version, dated, with what it changed over the one before.
- MiMo-V2-FlashEarlier, not in Sub Rosa
- 309B total and 15B active parameters, 256K context, hybrid sliding-window attention and multi-token prediction
- 73.4% on SWE-bench Verified, per Xiaomi
- MiMo-V2.5
- Native omnimodal encoders: vision and audio input on the V2-Flash backbone
- Context up to 1M tokens and post-training focused on agent work
- MiMo-V2.6-FlashLatest
- Higher Intelligence Index than V2.5-Pro (38 vs 26) at 309B total and 15B active
- 1M context, text and image input, very low price
Benchmarks
How to read these scores: what a benchmark tells you.
Artificial Analysis Intelligence Index: #32 of 90 models measured in the catalog. One score averaging about ten independent tests of reasoning, knowledge, math, coding and agent work, all run the same way for every model. Higher is better. Each model is listed at its highest reasoning effort. Two or three points apart is a close call; ten points is a different class of model.
See the numbers
| Model | points |
|---|---|
| 1. Claude Opus 5.5 | 58 |
| 2. Claude Sonnet 5.5 | 56 |
| 3. Claude Fable 5.1 | 53 |
| 4. GPT-6 Astra | 53 |
| 5. Gemini 4 Argon | 53 |
| 6. GPT-6.1 Sol | 52 |
| 7. Muse Wonder | 52 |
| 8. Claude Opus 5 | 51 |
| 9. Claude Fable 5 | 50 |
| 10. GPT-6 Sol | 48 |
| 32. MiMo-V2.6-Flash | 38 |
All 2 models in this family
| Model | Use | Privacy | Price |
|---|---|---|---|
MiMo-V2.6-Flashxiaomi-mimo-v2-6-flash | Text | Private | $0.157 per million tokens written$0.00011 a page written |
MiMo-V2.5xiaomi-mimo-v2-5 | Text | Private | $0.899 per million tokens written$0.000629 a page written |
Specifications
| Specification | MiMo-V2.6-Flash | MiMo-V2.5 |
|---|---|---|
| Released | September 21, 2026 | April 2026 |
| Parameters | 309B total, 15B active | 310B total, 15B active |
| Weights | Open, MIT | Open, MIT |
| Accepts | text, images | text, images, audio, video |
| Produces | text | text |
| Context | 1,000K tokens | 1,000K tokens |
| Longest answer | 131K tokens | 66K tokens |
| Abilities | Reads images, Uses Sub Rosa’s tools, Thinks before answering, Tuned for code | Reads images, Uses Sub Rosa’s tools, Thinks before answering, Tuned for code |
| Price | $0.157 per million tokens written | $0.899 per million tokens written |
| Privacy | Private | Private |
Prices, context and lengths from the live catalog on October 16, 2026; parameters and licenses from the maker.
Sources
Everything on this page comes from these pages, read on October 16, 2026 or as dated on each score.
Show the 9 sources
- artificialanalysis.ai/models/mimo-v2-flash
- huggingface.co/XiaomiMiMo/MiMo-V2.5
- artificialanalysis.ai/models/mimo-v2-6-flash
- artificialanalysis.ai/models/mimo-v2-6-pro
- artificialanalysis.ai/models/mimo-v2-5-pro
- mimo.xiaomi.com/
- api.venice.ai/api/v1/models
- openrouter.ai/xiaomi/mimo-v2.6-flash
- huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL