MiMo-V2.6-Flash
Xiaomi MiMo🇨🇳 CNMajor lab· Sep 21, 2026
A model variant within the MiMo-V2.6 family using a sparse Mixture-of-Experts architecture with 15 billion active parameters out of 309 billion total. Processes text, images, video, and audio in one model. Trained with groupwise reinforcement learning designed to expand capabilities through self-improvement — targeting complex agentic tasks, tool use, and multi-session long contexts.
Summary of the model card
- Context
- 1M tokens
- Price / 1M tok
- $0.14 in · $0.28 out
- License
- mit
- Modalities
- text, image, video, audio → text
2 variants
Sources
HuggingFaceOpenRouterBlogDeutsch
Eine Modellvariante der MiMo-V2.6-Reihe mit sparcer Mixture-of-Experts-Architektur (15 Milliarden aktiv pro Schritt von 309 Milliarden gesamt). Verarbeitet Text, Bilder, Video und Audio in einem Modell. Schulung mit selbstverbesserndem Gruppen-Verstärkungslernen für komplexe Agentenaufgaben, Tool-Nutzung und Multimodalität über lange Kontexte.
Provenance
- Detected
- Sep 27, 2026
- First source
- openrouter
- Description
- Summary of the model card
- Editorially reviewed
- —