Post-Cutoff
Search For AI agents

MicrosoftMAI-Image familyPreviewImage generationReleased 4 Sep 2026

MAI-Image-2.6

Microsoft AI’s flagship in-house image model (diffusion-based), for generation and controllable editing, with Bing grounding.

Model idMAI-Image-2.6

Price$5 text in, $38 image out per 1M tokens

Checked25 October 2026

Where it stands

6newer image generation models since its release

The 5 newest are listed. All image generation models

Weights
Closed weights; access is hosted only.

How to call it

4 ways to use MAI-Image-2.6, with the model id each provider expects.

Providers, model ids, endpoints and docs for MAI-Image-2.6
ProviderModel idEndpointDocs
Microsoft Foundry (Foundry Models sold by Azure) MAI-Image-2.6 https://<resource-name>.services.ai.azure.com/mai/v1/images/generations learn.microsoft.com
Microsoft Foundry (image edits) MAI-Image-2.6 https://<resource-name>.services.ai.azure.com/mai/v1/images/edits learn.microsoft.com
OpenRouter microsoft/mai-image-2.6 https://openrouter.ai/api/v1/images openrouter.ai
Web app (MAI Playground) playground.microsoft.ai

Pricing

Prices for MAI-Image-2.6
Text input$5 per 1M tokens
Image input$8 per 1M tokens
Image output$38 per 1M tokens

As published: “USD per 1M tokens (OpenRouter listing, provider Azure); Artificial Analysis estimates ~$38.9 per 1,000 images at 1024x1024”.

Source: openrouter.ai, checked 25 October 2026.

Specs

Input
Text and images
Output
Images
Context window
Not published
Max output
Not published
Open weights
No
Licence
Proprietary
Released
4 September 2026
Status
Preview
Type
Image generation

What stands out

What its maker marketed at launch, and what people found later, each with its source.

  • Bing web grounding

    web_grounding: true lets the model retrieve current information from Bing Search and use it as context for real-world entities, places and events (MAI-Image-2.6 and -Flash only).

    Source: learn.microsoft.com

  • Model-chosen aspect ratio

    auto_aspect_ratio: true lets the model pick the output aspect ratio from the prompt and input images; output up to 2,359,296 px total (1536x1536 equivalent), minimum 768 px per side.

    Source: learn.microsoft.com

  • Multi-image reference editing

    Edits API takes up to five JPEG/PNG reference images and combines people, products, styles and scenes; precise local edits (object removal, inpainting, text updates) that preserve layout.

    Source: microsoft.ai

Notes

Artificial Analysis text-to-image leaderboard (read 2026-10-25): #5, Elo 1166, listed release Aug 2026, $38.9 per 1k images (https://artificialanalysis.ai/text-to-image/arena/leaderboard-text). AA image-editing leaderboard: tied 3-6, Elo 1151 (https://artificialanalysis.ai/image/leaderboard/editing). Foundry model version 2026-07-31, Global Standard deployment, public preview; PNG output only; English prompts; 32k-token prompt limit; rate limits 6-36 RPM by tier. OpenRouter versioned id microsoft/mai-image-2.6-20260904. No official Azure price page was found: token prices are from OpenRouter’s Azure-provider listing. Photorealistic edits of minors blocked by default.

curl -X POST "https://$RESOURCE.services.ai.azure.com/mai/v1/images/generations" \
  -H "Content-Type: application/json" -H "api-key: $AZURE_API_KEY" \
  -d '{"model":"<your-deployment-name>","prompt":"A poster of a lighthouse at dawn","width":1024,"height":1024,"web_grounding":false}' \
| jq -r '.data[0].b64_json' | base64 --decode > output.png

Sources:

Not yet checked against an official source: Some secondary sources date a first announcement to 2026-08-10 (unverified).