Models·Xiaomi

MiMo-V2.6-Pro model review

xiaomi/mimo-v2.6-pro

Xiaomi's open-weight model for text, images, video and audio, with a very large context window and tool support.

$0.43Input, per million tokens
$0.87Output, per million tokens
1.05MContext window
131KLongest answer

1.05M tokens of context — about 787,500 words, on a scale that ends at 2M.

What it can do

  • Reasoning
  • Vision
  • Tool use
  • Open weights
  • Takes text
  • Takes image
  • Takes video
  • Takes audio
  • Gives text

Released 21 Sept 2026 · checked 3 Oct 2026.

About it

MiMo-V2.6-Pro is Xiaomi's open-weight foundation model, released September 21, 2026. It accepts text, images, video and audio, then returns text. The model supports reasoning and tools, can read up to 1,050,000 tokens at once, and can produce up to 131,072 tokens in one answer. Its listed prices are $0.435 per million input tokens and $0.87 per million output tokens.

Review

In a line

MiMo-V2.6-Pro is Xiaomi's foundation model for working with text, images, video and audio. It is open weight, so the model files are available for people who want to run or adapt them.

What it's good at

The model is built for tasks that need reasoning and can use tools. It takes more than text as input, including images, video and audio, while its responses are text.

Its context window is unusually large, and its design uses more than one trillion parameters. Those details suit long documents or projects that need several kinds of input.

Price in plain terms

Input costs $0.435 per million tokens, and output costs $0.87 per million tokens. In other words, generated text costs twice as much as text sent to the model.

Using the rough examples supplied, reading a 500-word page costs about 0.028 cents. One chat turn with about 1,000 input tokens and 500 output tokens costs about 0.087 cents. One thousand such turns cost about $0.87.

Context and limits

MiMo-V2.6-Pro can take up to 1,050,000 tokens in one request. Tokens are small pieces of text, so this is enough room for very long material, along with other supported inputs.

A single answer can contain up to 131,072 tokens. The model returns text only, even though it accepts text, image, video and audio inputs.

Ways to use it

The model's open weights allow use in software that supports downloadable model weights. Its tool support is useful in apps or chat tools that let a model call tools.

It is identified as xiaomi/mimo-v2.6-pro. The related model files are listed as XiaomiMiMo/MiMo-V2.6-Pro-RL. People comparing it with other choices can browse Xiaomi models, open-weight models, or newest models.

Verdict

Choose MiMo-V2.6-Pro if you need one model that can accept several media types, handle very long context, reason through tasks and use tools. Its open weights also suit projects that need more control over deployment.

Look elsewhere if you need an answer in an image, video or audio format, because this model outputs text. It may also be more capacity than a short, simple text task requires.

Verdict

MiMo-V2.6-Pro is a sensible pick for teams handling long documents or mixed media while keeping model files available for their own setup. The per-token rates are low in the supplied examples, including about $0.87 for 1,000 typical chat turns. Its main trade-off is output format: every response is text, so it cannot directly deliver images, video or audio. Short tasks may not benefit from its million-token context window.

  • Accepts text, image, video and audio inputs
  • Open weights are available
  • Supports reasoning and tool use
  • 1,050,000-token context window
  • Up to 131,072 tokens in one response
  • Responses are text only
  • Requires software that supports its open weights
  • Large context may be unnecessary for short tasks
  • Pick it ifLong documents and mixed-media input; Apps that can connect models to tools; Teams wanting open model weights
  • Skip it ifYou need generated images, video or audio; Your software cannot run open-weight models; You only handle short, simple prompts

Pricing

Input is $0.435 per million tokens and output is $0.87 per million tokens. The supplied examples put 1,000 typical chat turns at about $0.87.

Questions

Who makes MiMo-V2.6-Pro?

Xiaomi developed and released MiMo-V2.6-Pro on September 21, 2026.

Is MiMo-V2.6-Pro open source?

Its weights are open. The listed Hugging Face identifier is XiaomiMiMo/MiMo-V2.6-Pro-RL.

What can I send to the model?

It accepts text, images, video and audio as inputs.

What does it return?

The model returns text.

How much context can it read?

Its context window is 1,050,000 tokens, meaning one request can include a very large amount of material.

What is the maximum answer length?

One response can contain up to 131,072 tokens.

How much does a typical chat cost?

A turn using about 1,000 input tokens and 500 output tokens costs about 0.087 cents. One thousand such turns cost about $0.87.

On these boards

More from Xiaomi