MiMo-V2.6-Flash
Budgetby Xiaomi
MiMo-V2.6-Flash is the low-cost half of Xiaomi's MiMo-V2.6 release on September 21, 2026, published under MIT as XiaomiMiMo/MiMo-V2.6-Flash-RL. It is a 309 billion parameter mixture-of-experts model activating about 15 billion per token, with the same native text, image, video, and audio input and 1 million token context as the Pro model. At $0.14 per million input tokens, $0.28 output, and $0.0028 cached, unchanged from V2.5, it is one of the cheapest rate cards on this site for a model that takes all four input types. The surprising number is independent: the Vals Index puts Flash at 59.58 percent, fractionally ahead of its own Pro sibling at 59.47 and ahead of DeepSeek V4.1 Flash at 57.86. On Xiaomi's table it trails Pro by a few points on most agent rows, 67.9 against 71.9 on DeepSWE v1.1 and 87.6 against 89.9 on Terminal-Bench 2.1, and falls further behind on Terminal-Bench 4.0 at 28.8 and on the cybersecurity exploitation rows. Xiaomi says Flash completed its reinforcement learning run for about $850,000 in under six days, which is a useful data point for anyone pricing their own post-training. All benchmark figures other than the Vals and Artificial Analysis results are vendor-reported.
Input Price
$0.14
per 1M tokens
Output Price
$0.28
per 1M tokens
Context Window
1.0M
tokens
Released
2026-09
Open source
Capabilities
Key Strengths
- ✓MIT-licensed open weights at 309B total and 15B active parameters
- ✓$0.14/$0.28 per 1M tokens with cached input at $0.0028
- ✓Vals Index 59.58 percent, level with the Pro model
- ✓Text, image, video, and audio input in a 1M token context
- ✓Terminal-Bench 2.1 at 87.6, vendor-reported
Best For
- ▸High-volume omnimodal pipelines on a tight budget
- ▸Sub-agents and routing tiers under a larger planner model
- ▸Self-hosted coding assistants on a smaller cluster than the Pro model needs
- ▸Batch video and audio understanding
Benchmark Scores
| Benchmark | Score | Description |
|---|---|---|
| Terminal-Bench 4.0 | 21.2 | Long-horizon agentic work in a terminal across software, science, ML, operations, hardware, security, and media (66 tasks, all-or-nothing verifiers) |
Scores sourced from public benchmark datasets. See full benchmark leaderboard for all models.
Pricing Details
Input tokens
$0.14
per 1M tokens
Output tokens
$0.28
per 1M tokens
Estimated cost per 1K requests
$0.28
~1K input + ~500 output tokens avg
Prices are subject to change. Check the official documentation for current pricing. See the cost calculator for detailed estimates.
Open Source Model
MiMo-V2.6-Flash is free to download and self-host under the MIT. Hosted API pricing varies by provider (e.g., Together, Fireworks, Groq). See our open source LLM guide for deployment options.