Microsoft Launches MAI-Image-2.7-Preview for Controlled Image Editing, Ranks Second Globally

Microsoft released MAI-Image-2.7-Preview on October 27, an image editing model designed for controllable edits using an input image plus text instructions. Text-to-image generation is not yet available, and the model is positioned as a tool for editing existing photos, design drafts, screenshots, and film material.

According to Artificial Analysis, the model scored 1,193 Elo, ranking second on the image editing leaderboard, behind OpenAI's GPT Image 2.5 Sunburst (max) and ahead of offerings from Google, Meta, and xAI. It leads the next best model by 24 Elo in scene and style editing.

Microsoft said that compared with the previous MAI-Image-2.6, the new version improved across all 10 test cases and 7 editing action categories, reaching category-leading performance in enhancement/restoration, reasoning-based editing, scene style, and text/symbol handling. Users can currently try it on Arena, Artificial Analysis, and MAI Playground. Text-to-image is said to be "coming soon," but no date has been announced.

The model remains a preview release. Microsoft has not yet disclosed public API access, pricing, or general commercial availability, and model weights are not open for self-hosting.