ModelRadar
/

MiMo-V2.6-Pro

Xiaomi MiMo🇨🇳 CNMajor lab· Sep 21, 2026

TextReasoningVisionopen weightsAPI

Xiaomi MiMo-V2.6-Pro is a multimodal language model natively handling text, images, video, and audio with a 1M-token context window. It is trained via asynchronous reinforcement learning combining code generation, general-purpose agents, visual understanding, and cybersecurity into a single mixed-task training run.

Summary of the model card

Context
1.1M tokens
Price / 1M tok
$0.43 in · $0.87 out
License
mit
Modalities
text, image, video, audio → text

2 variants

Standard1.1M$0.43/$0.87OR
RLHF

Sources

HuggingFaceOpenRouterBlog
Deutsch

Xiaomi MiMo-V2.6-Pro ist ein multimodales Sprachmodell, das Text, Bilder, Video und Audio原生 unterstuetzt und ein Kontextfenster von 1 Million Token bereitstellt. Es wurde mit asynchronem Reinforcement Learning trainiert, das Codierung, allgemeine Agents, visuelles Verstaendnis und Cybersicherheit in einer gemeinsamen Aufgabe kombiniert.

Provenance

Detected
Sep 27, 2026
First source
openrouter
Description
Summary of the model card
Editorially reviewed
—