Xiaomi MiMo

What it is

Published price

• WeightsMIT license, commercial use and fine-tuning allowed$0
• API — MiMo-V2.6-Pro$0.435 in / $0.87 out per million tokens

The published figures for Xiaomi MiMo, copied exactly as their sources give them and checked 17 September 2026. Each one names the test, who ran it and when, and says when the number comes from the maker rather than an independent tester.

Where to read today's numbers

Scores move, so these are the boards that keep them current.

• Hugging Face open model leaderboardsIndependent

  Standardized test scores for open weights you can download and run yourself.

• Artificial AnalysisIndependent

  Independent testing of quality, speed and price per million tokens across hosted models.

• Xiaomi's own model cardsMaker's own figures

  The maker's published scores, alongside the weights. Compare against the open leaderboards.

How it performs, in my experience

Solid open reasoning models. Nowhere near the biggest hosted models, plenty for summarizing, drafting and code help.

• Free value

  5 / 5

  How much you get without paying.

• Performance

  3 / 5

  How well it does its main job.

• Range of uses

  3 / 5

  How many different jobs it suits.

Free open weights you can run privately; slower than anything hosted. These scores are my own judgment, not a measurement — the benchmark links above are the independent version.

Prompting it from this site

Endpoint wired, waiting on a key

MiMo V2.5 through OpenRouter, about $0.14 per million tokens in — the cheapest of the four, though the card now tracks the newer MiMo-V2.6-Pro. Needs an OpenRouter key.

Use it for

• Running a capable model on your own hardware
• Work you don't want sent to anyone's server

Runs on your own machine

MiMo's weights are published openly, and the smaller sizes are runnable on a single decent GPU — this is one of the more approachable open models here.

What you need first

• A machine with an NVIDIA GPU (check the model card for the memory needed by the size you choose)
• Python 3.10+ with a working CUDA install

Step by step

1. 1.Install the tooling

   pip install vllm transformers huggingface_hub

2. 2.Download the weights

   hf download XiaomiMiMo/MiMo-7B-RL --local-dir ./mimo

3. 3.Serve it locally

   vllm serve ./mimo --trust-remote-code

For a CPU-only machine, look for a community GGUF conversion and run it with Ollama or llama.cpp instead — slower, but it works without a GPU.

Its own pages

• Xiaomi MiMo — main site
• Xiaomi's MiMo pricing docs
• MiMo weights on Hugging Face
• MiMo documentation