MAI-Image-2

Name: MAI-Image-2
Author: Microsoft Research

ProprietaryMultimodal

About

Microsoft's most capable image generation and understanding model. Supports both image generation and vision understanding tasks. Available in Microsoft Foundry and MAI Playground for developers with enterprise-grade capabilities.

Capabilities

VisionMultimodalReasoningFunction CallingTool UseJSON ModeCode Execution

Rankings

Best Multimodal / Vision LLMs Cheapest LLM APIs

Specifications

FamilyMAI

Released2026-04-02

ArchitectureTransformer

SpecializationImage Generation and Vision Understanding

TrainingSupervised learning on image and vision data

Created by

Microsoft Research

Advancing the state-of-the-art in AI and computing.

Redmond, Washington, United States

Founded 1991

Website

Links

Website