MiMo-V2.6-Pro
Xiaomi's flagship open research model under a permissive MIT license, balancing strong reasoning and coding with a notably low served price.
At a glance
Released
May 14, 2026
Provider
Xiaomi
Context window
256,000 tokens
Max output
32,000 tokens
Modalities
Text + image → text
Knowledge cutoff
January 2026
API model ID
mimo-v2.6-proLicense
Open weights (MIT)
Pricing
Standard
Input$0.435 / 1M
Output$0.87 / 1M
* Rates are from a third-party aggregator — verify before relying on them. A typical 1,500-token prompt with a 650-token answer costs ≈ $0.001 on the standard tier. Run your own prompt through the calculator →
What's new vs the previous model
Compared withMiMo-V2.1-Pro (January 2026)
- Improved coding and reasoning accuracy
- Added image understanding and longer context
- Higher memory needs than the previous generation
Best use cases
Coding assistants
Generates and reviews code with reliable instruction following.
Permissive commercial use
MIT licensing makes it easy to embed in products without legal friction.
Research fine-tuning
Serves as an open base for academic and industrial experiments.
Strengths & limitations
Strengths
- Fully permissive MIT license
- Strong coding
- Low cost
Limitations
- Smaller ecosystem than Llama
- Modest multimodal breadth
How to test it
- Download the weights: grab MiMo-V2.6-Pro from Hugging Face (XiaomiMiMo) ↗.
- Run it locally: serve the open weights with vLLM, SGLang or Ollama on your own GPU.
- Call it: hit the local OpenAI-compatible endpoint it exposes (example below).
# serve the open weights, then call the local OpenAI-compatible endpoint
vllm serve XiaomiMiMo/mimo-v2.6-pro
curl http://localhost:8000/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{"model": "mimo-v2.6-pro", "messages": [{"role": "user", "content": "Explain tokenization in one paragraph."}]}'