MAI Image 2.7: Microsoft's new AI image editor, explained
Microsoft's MAI Image 2.7 preview edits photos: relight scenes, change text, move products. Where it ranks, how to try it and what's still missing.
By Olena, me20 Studio · 2 min read

Microsoft AI released MAI Image 2.7 in preview on 27 October 2026. It is an editing-only model for now: you upload a picture and describe the change — relight a scene, swap text on a label, move a product into a new setting. It debuted at No. 3 on Arena's image-edit leaderboard and No. 2 on Artificial Analysis.
What is MAI Image 2.7?
MAI Image 2.7 is Microsoft's newest in-house image editing model, the next step after MAI-Image-2.6. Microsoft says it starts with editing, with text-to-image generation "coming soon"; in the MAI Playground, 2.7 is offered for edits while generation still uses 2.6.
Microsoft lists the edits it handles best:
- restyling or relighting a whole scene
- changing text and symbols so they still look printed into the image
- restoring and enhancing old or low-quality assets
- moving a product into a new setting, shifting light and mood, or reworking one object without disturbing the rest
How does it rank against GPT Image 2.5?
MAI Image 2.7 sits in the top group of editing models, next to OpenAI's GPT Image 2.5. On the Artificial Analysis editing leaderboard (28 October), GPT Image 2.5 Sunburst scored 1199 Elo, MAI Image 2.7 1193 and GPT Image 2.5 Flare 1182 — but the confidence ranges overlap, so the order of the top three is not decisive.
| Leaderboard | MAI Image 2.7 | Ahead of it | |---|---|---| | Artificial Analysis (editing) | No. 2 | GPT Image 2.5 Sunburst | | Arena (image edit) | No. 3 | GPT Image 2.5 Sunburst and Flare (commercial-design board) |
Artificial Analysis also put 2.7 first in four editing actions: scene and style, text or symbols, reasoning-based edits, and enhancement and restoration.
What is still missing?
The preview has no announced price, no public API and no date for Microsoft Foundry, where earlier MAI image models reached developers. Text-to-image generation for 2.7 has no date either.
How to try it on product images
Open the MAI Playground, upload a product photo you already approved and run three edits you actually need: a new background and light, a text change on the packaging, and a clean-up of an old asset. Then run the same three edits in another editor, such as Krea or ChatGPT Images, and compare side by side.
- check that the logo, colour and shape of the product did not change
- read every letter of edited text
- for print or large banners, upscale the final edit in Magnific
More on product edits: our Krea 3 editing guide, lifestyle vs packshot images and our imagery portfolio. Need on-brand AI visuals for your products? Tell us about your project.
P.S. The caption under the cover image is a ready-to-use prompt. Copy it into Midjourney, Higgsfield or any image generator and make your own version.