Microsoft AI's image editing model, now in preview
MAI Image 2.7 is Microsoft AI's latest image editing model, released as a preview on October 27, 2026. It changes images you already have: restyle or relight an entire scene, rewrite text and symbols so they stay part of the picture, restore and enhance existing assets, and make edits that depend on how objects, light and environments interact. Microsoft reports that it debuted at No. 2 on Artificial Analysis's image editing leaderboard and No. 3 on Arena Edit.
The preview can be tried on MAI Playground, Arena and Artificial Analysis. Microsoft has not yet released a public API, pricing or text-to-image generation for MAI Image 2.7. PoYo will add it here once an API is available; until then, MAI Image 2.6 and MAI Image 2.6 Flash already provide text-to-image, single-image editing and multi-reference composition through the PoYo API.
Change the style, time of day or lighting of a whole image from one text instruction while keeping its layout and subjects. Scene and style editing is Microsoft's strongest reported result: 24 Elo ahead of the next-best model on Artificial Analysis.
Rewrite words on signs, labels and packaging, or swap icons and symbols, while keeping them integrated with the surface, perspective and lighting around them. Microsoft places MAI Image 2.7 at the category frontier for text and symbol editing.
Repair aged photos and improve low-quality brand images instead of generating new ones, so the original subject and composition stay recognizable. Microsoft reports frontier-level results for enhancement and restoration on Artificial Analysis.
Move a product into a new setting or rework one object without breaking the rest of the scene. MAI Image 2.7 is built for edits that depend on how objects, lighting and environments work together, with shadows and light matched to the result.
Place one product photo into new seasonal or lifestyle settings and relight it to match, so catalog teams can produce campaign variations with MAI Image 2.7 without booking a new shoot for every scene.
Update headlines, offers and labels on existing ad creatives while keeping the layout and lighting intact. MAI Image 2.7's text and symbol editing suits localized campaigns and quick promotional refreshes.
Change the time of day, mood or lighting of a finished photo with a written instruction. MAI Image 2.7 relights and restyles whole scenes, giving retouchers a faster first pass before detailed manual work.
Restyle rooms, swap furniture or show a property in different lighting from a single photo. Reasoning-based edits in MAI Image 2.7 help keep objects, shadows and perspective consistent with the space.
Restore and enhance aged photos and legacy visual assets for reuse in articles, books and archives. MAI Image 2.7 improves existing images rather than inventing new ones, so the original subject stays recognizable.
Plan photo-editing, design and creative apps around MAI Image 2.7. Build the workflow today with MAI Image 2.6 on PoYo, then evaluate 2.7 against your own test images once its API is released.
MAI Image 2.7 is Microsoft's editing-focused preview; MAI Image 2.6 is the current production model for both generation and editing. MAI Image 2.7 details come from Microsoft AI's October 27, 2026 announcement; MAI Image 2.6 limits are the parameters PoYo exposes today.
| Feature | MAI Image 2.7 | MAI Image 2.6 |
|---|---|---|
| Release | Preview, announced October 27, 2026 | Announced August 10, 2026 |
| Status on PoYo | Coming soon (no public API yet) | Available now |
| Focus | Image editing | Text-to-image and image editing |
| Text-to-image | Not yet (Microsoft says coming soon) | Yes |
| Editing strengths | Scene restyling and relighting, text and symbol edits, restoration, reasoning-based edits | Single-image edits and multi-reference composition |
| Editing benchmark | No. 2 on Artificial Analysis and No. 3 on Arena Edit at launch; ahead of 2.6 in all 10 use cases and 7 editing actions | Scores below MAI Image 2.7 in every Artificial Analysis editing category |
| Reference images | Not yet published | Up to 5 |
| Web grounding | Not yet published | Yes |
| Maximum output | Not yet published | 2,359,296 pixels (1536×1536 equivalent) |
Step 1: Register & Create API Key
Sign in to PoYo and create an API key. The same key will work for MAI Image 2.7 once it is added, and it already works for MAI Image 2.6, MAI Image 2.6 Flash and every other model on PoYo.
Create API Key →
Step 2: Top Up Credits
Add credits to your account. MAI Image 2.7 pricing will be published on this page when the API is available; current image models are billed by usage, with failed tasks not charged.
Add Credits →
Step 3: Build Your Editing Workflow Today
Start with MAI Image 2.6: send a prompt plus 1 to 5 images in image_urls to edit a single image or combine several. When MAI Image 2.7 becomes available, its model ID, parameters and pricing will be listed here and in the API documentation.
View MAI Image 2.6 Documentation →
MAI Image 2.6 and MAI Image 2.6 Flash already run on PoYo with one request format. MAI Image 2.7 will be added to the same API and key, so moving to the new model will not require a new provider, a new account or a separate billing setup.
Image models on PoYo show their rates and an estimated cost before you submit, with no subscription and no charge for failed tasks. MAI Image 2.7 pricing will be published on this page as soon as Microsoft releases the API and the model is added to PoYo.
Test prompts, reference images and output sizes for MAI Image 2.6 in the browser today, then copy the JSON request straight into your code. MAI Image 2.7 will get the same Playground, so you can compare results before switching models in production.
Evaluate GPT Image 2.5, Nano Banana 2.1, Seedream 5.0 Pro and MAI Image 2.6 side by side with one API key while you wait for MAI Image 2.7, and keep the model that best fits your images, budget and quality bar.
