What Is MAI Image 2.6?
MAI Image 2.6 is Microsoft’s image generation and editing model, with improvements in lettering, portraits and commercial visuals. Use it to draft a packaging concept, compose a poster with a short headline, or describe a focused change to an existing photo. This page provides text-to-image generation and single-image editing with PNG downloads. The model remains in public preview.
Sources: Microsoft MAI Image 2.6 model card · September 4, 2026 · Microsoft image API documentation · checked September 6, 2026
MAI Image Model Versions
MAI Image 2.6
Public preview · 1K PNG
From 5 Credits
MAI Image 2.6
Microsoft’s image model for readable poster text, natural portraits and targeted photo edits.
- Resolutions: 1K
- Reference images: up to 1
- Modes: Text to Image, Edit Image, Typography
- Speed: Balanced
MAI Image 2.5 Pro
Microsoft’s flagship MAI image model for precise typography, rich detail and production-ready compositions.
- Resolutions: 1K
- Reference images: not supported
- Modes: Text to Image, Typography
- Speed: Slow
MAI Image 2.5
Microsoft’s photoreal image model for natural people, polished materials and design-ready visual concepts.
- Resolutions: 1K
- Reference images: not supported
- Modes: Text to Image, Typography
- Speed: Balanced
What Is MAI Image 2.6 Best For?
- Posters and packaging with short, quoted text
- Product concepts with controlled lighting and materials
- Natural editorial portraits
- Changing an object, background or label in an existing photo
How to Use MAI Image 2.6
- Choose Text to Image or Edit Image.
- For editing, upload one PNG or JPEG.
- Describe the subject, lighting and composition in up to 1,000 characters; quote any required text.
- Choose an aspect ratio for a new image and check the credit total.
- Generate, proofread lettering and inspect important details before downloading.
MAI Image 2.6 Settings Explained
Mode
Text to Image starts from your description. Edit Image requires one PNG or JPEG and a description of what to change and what to preserve.
Aspect Ratio
For a new image, choose square, landscape or portrait before generating. Editing uses the model’s framing; the aspect-ratio selector is hidden in that mode.
Credits
A new 1K image costs 5 credits. Editing one image costs 7 credits. One request produces one PNG, and failed generations are refunded.
The panel above shows only the parameters MAI Image 2.6 actually supports. Switch versions and unsupported controls disappear rather than sitting there doing nothing.
MAI Image 2.6 Prompt Examples
A poster with exact copy
Design a portrait poster for a fictional ceramics fair. Show three handmade bowls on warm ivory linen. A large dark green headline reads exactly “FORM & EARTH”. A smaller line reads “SATURDAY 12 SEPTEMBER”. Keep the lower fifth empty. Soft daylight, crisp lettering, no other text.
A product material study
Studio photograph of a matte sage-green insulated bottle on pale limestone. A small label reads exactly “FIELD”. Keep the bottle upright and centered, with a soft shadow falling right. Show realistic brushed metal on the lid and subtle condensation. Clean warm background.
A focused photo edit
Replace the background with a softly lit cream studio wall. Preserve the person’s face, hairstyle, clothing, pose and proportions. Keep the direction of the original light. Do not add text or objects.
Select Edit Image and upload one PNG or JPEG.
MAI Image 2.6 Test Results
Testing in progress
Every model runs the same eight-category prompt set — portrait, product, poster text, anime character, interior, multi-reference, image editing and character consistency — recording generation time, success rate, text accuracy and identity retention. That run has not finished for MAI Image 2.6, so this page shows no scores. We do not publish numbers we have not measured.
MAI Image 2.6 Limitations
- One output per request; prompts up to 1,000 characters
- Current output tier: 1K, with a maximum of 1,048,576 pixels
- Editing accepts one PNG or JPEG; the model determines the edited image framing
- Multi-reference editing, web grounding and 1.5K output are not enabled here yet
- Public preview availability and speed can vary; in-image text still needs proofreading
MAI Image 2.6 Pricing and Credits
Credits scale with resolution and image count. The Generate button always shows what this run costs.
1K
5 credits / image
Failed generations cost 0 credits.
Commercial Use and Image Rights
Commercial use is subject to Microsoft’s applicable service terms, including the terms for public preview features. Review generated text, logos and product details before publication.
- Model used: MAI Image 2.6
- Model developer: Microsoft AI
- The inference route is managed internally and may change without changing the selected model
- Uploads are sent only to the configured inference service needed to complete the generation
- You can delete any generation from your history at any time
Frequently Asked Questions
Is MAI Image 2.6 the same as MAI Image 2.5 Pro?+
No. These are separate model versions. Selecting 2.6 requests 2.6; the site does not replace it with 2.5 or 2.5 Pro when it is unavailable.
How many credits does MAI Image 2.6 use?+
Text-to-image and typography generation cost 5 credits per 1K image. A single-image edit costs 7 credits. Failed generations cost 0 credits.
Is MAI Image 2.6 better than GPT Image 2?+
That depends on the task. MAI Image 2.6 is a useful candidate for lettering, portraits and focused edits. GPT Image 2 currently offers more reference images and larger output sizes on this site. We have not completed a controlled comparison, so we do not claim a quality winner.
Why is output limited to 1K when Microsoft mentions 1.5K?+
Microsoft’s September 2026 model card describes up to 1.5K output, while the current image API documentation still specifies a 1,048,576-pixel limit. This page uses the documented API limit until larger outputs are verified.
Can I use multiple references or web search?+
Those capabilities are described for the wider model release. This page currently offers one-reference editing and does not enable web search.
What does Coming soon mean?+
The introduction and settings are available, but generation is disabled on deployments where this model has not been configured. An unavailable request does not consume credits.
Release History and Sources
August 10, 2026 · Model announced
August 10 is the first model announcement. The September Foundry public preview is a separate platform-availability event.
Explore this event on the image model timeline →
Related Models and Tools
AI Image Generator
Generate with all supported image AI models. Compare images from one prompt.
Poster Generator
Posters with headlines that come out right the first time.
Social Post Generator
One visual, every platform size, one consistent style.
GPT Image 2.5 Flare
OpenAI
Create product photos, posters and image assets with OpenAI’s faster Image 2.5 model.
GPT Image 2.5 Sunburst
OpenAI
Refine product photos and reference images with OpenAI’s most capable Image 2.5 editor.
GPT Image
OpenAI
Generate and edit images with detailed prompts, multiple reference photos and readable text.
GPT Image 1.5
OpenAI
OpenAI’s high-fidelity image generation model for precise composition, lighting, detail and readable text.