Mage

A Lightweight, Research-Friendly Multimodal Model Family

Microsoft Mage Team
Code

🤗

Models
Mage family

Mage is a family of lightweight, research-friendly multimodal models built at a 4B-parameter scale, spanning visual understanding, generation, and computer use. Mage-VL and Mage-Flow share a codec-aligned efficiency philosophy—spend representation capacity where the signal is. Mage-VL-Cowork extends the family to desktop workflows, learning to combine graphical interaction and programmatic execution in one compact policy.

Vision–Language · Understanding
Mage-VL
An Efficient Codec-Native Streaming Multimodal Foundation Model
A codec-native, from-scratch VLM for image & video understanding — reads video the way a codec does (anchor/predicted frames, 16×16 patches), with bio-inspired proactive streaming.
Enter project page → Generation · Editing
Mage-Flow
An Efficient Native-Resolution Foundation Model for Image Generation and Editing
A compact 4B generative stack (Mage-VAE + Native-Resolution MMDiT) for text-to-image generation and instruction-based editing at native resolution, with Base / RL / 4-step Turbo variants.
Enter project page → Computer Use · GUI + Code
Mage-VL-Cowork
A Full Training Recipe to Learn Hybrid Computer Use from Evolved Harnesses
A 4B hybrid computer-use policy that learns when to click and when to code. Mage-CUA-Forge, evolved teacher harnesses, and a progressive five-stage curriculum deliver 66.0% on OSWorld-Verified under a 50-step budget.
Enter project page →