Image

MAI-Image-2.6 API

MAI-Image-2.6 API access to Microsoft AI's latest image model. Call it with one APIMart API key, with no Azure or Foundry deployment to set up. Generate from text, edit one image or combine up to five references, and get one PNG per request, billed by actual text-input, reference-image and output-image tokens.

Text to ImageUp to 5 References1K / 2KBing GroundingToken Billing

Get API Key API Docs

Service Highlights

Official Discounts
Pay-as-you-go
Fast & Low Latency

Input

Model

Prompt

Reference Image

Size mode

Resolution

Aspect Ratio

Bing search grounding

Run estimates image-output tokens only; text and reference-image input cost extra. The estimate is — when auto or reference-based dimensions are unknown.

Output

Generated image links are valid for 72 hours after completion. Save your images promptly.

MAI-Image-2.6 API Pricing

Transparent pricing with no hidden fees. Pay only for what you use.

Variant	Tier	GoldCurrent price	Platinum	Diamond	Official	Your savings
`mai-image-2.6` Prices are estimates; actual cost is billed by token usage.	1K · 1024×1024	0.3113Credits(~$0.0311)/pic	0.2957Credits(~$0.0296)/pic	0.2802Credits(~$0.028)/pic	0.3891Credits(~$0.0389)/pic	20%
	2K · 1536×1536	0.7004Credits(~$0.07)/pic	0.6654Credits(~$0.0665)/pic	0.6304Credits(~$0.063)/pic	0.8755Credits(~$0.0876)/pic	20%

* Actual costs are subject to final output.

* Image-output estimate only, using 1024×1024 for 1K and 1536×1536 for 2K. Text and reference-image input are billed separately; actual tokens determine the final charge.

Key Features of the MAI-Image-2.6 API

What Microsoft's model does well, and the controls the API exposes

Text, Portraits and 3D in Microsoft MAI-Image-2.6

Microsoft describes MAI-Image-2.6 as an upgrade over MAI-Image-2.5 in text rendering, portraits, 3D imagery and polished commercial and photorealistic output. Test with your own prompts before production use.

Text to Image with MAI-Image-2.6

Describe a photograph, poster or product scene. Include the composition, lighting and any exact text in quotes, then review the result before publishing.

Image Editing with Up to Five References

Pass one image in image_urls to edit it, or up to five to combine them. Use public image URLs or Base64 data URLs and describe what each reference should contribute.

1K and 2K Output from the MAI-Image-2.6 Model

For text-to-image, select a resolution tier and an aspect ratio from 1:4 to 4:1, or let the model choose with auto. 4K is not supported.

Exact Pixel Sizes in the mai-image-2.6 API

Set paired width and height, each at least 768 pixels, within 2,359,296 total pixels. Use multiples of 32; 3072×768 is valid.

Bing Search Grounding in the MAI Image 2.6 API

Enable web_grounding to retrieve current information with Bing before generating people, places or events. It is off by default.

Reference-Led Dimensions

For image editing, the model chooses the output dimensions from the references. Size, resolution, width and height settings do not control the result.

One PNG per MAI-Image-2.6 API Request

Each request returns one PNG. Billing includes actual text input, reference-image input and image-output tokens; failed tasks are fully refunded.

What You Can Build with the MAI-Image-2.6 API

Practical ideas for marketing, product and visual-content teams

Text-Rich Marketing Creative with MAI-Image-2.6

Create campaign drafts, posters and social graphics from a brief. Check text, layout and brand details before use.

Product and Branding Imagery

Explore polished product shots, packaging concepts, branded compositions and commercial artwork from concise creative directions.

Portrait Concepts from Microsoft MAI-Image-2.6

Explore portrait-led editorial, lifestyle and campaign concepts. Review the result for accuracy, consent and appropriate use.

3D Concepts and Mockups

Develop rendered-object studies, dimensional visual directions and 3D-style mockups during early creative exploration.

Photorealistic Storyboards with the MAI-Image-2.6 Model

Turn scene descriptions into photorealistic or cinematic frames for mood boards, pitch materials and previsualization.

Reference-Guided Edits with the MAI-Image-2.6 API

Combine up to five references for a product scene or design direction, describing which elements to retain or change.

How to Get a MAI-Image-2.6 API Key and Start Building

Create an account, add funds and generate your first image

Create an APIMart Account

→

Add Funds (Pay As You Go)

Top up your balance. Usage is billed per request by actual tokens, and the same balance works across all models.

→

Get Your MAI-Image-2.6 API Key

Create a key in the console and send it as a Bearer token with every request to the MAI-Image-2.6 API.

Get Key

MAI-Image-2.6 API Example: Text to Image and Image Editing

Call the mai-image-2.6 model on APIMart with one POST request. Generation is asynchronous: submit a task, then poll the task endpoint to get the MAI-Image-2.6 image URL.

1
Authenticate MAI-Image-2.6 API requests
Save the key you created above in the APIMART_API_KEY environment variable and send it as Authorization: Bearer <key> on every request.
2
Submit a MAI-Image-2.6 task
POST /v1/images/generations with model mai-image-2.6 and your prompt. Add size and resolution for text-to-image, or image_urls to edit. The response returns a task_id in data[0].task_id.
3
Get the MAI-Image-2.6 result
Poll GET /v1/tasks/{task_id} every 3–5 seconds until status is completed, then read the PNG URL from data.result.images[0].url[0]. Download it before expires_at (results are kept for 72 hours). A failed task returns the reason in data.error.message and is fully refunded.
See the MAI-Image-2.6 API reference for every parameter, size table and error code.

POST /v1/images/generations

curl -X POST https://api.apimart.ai/v1/images/generations \
  -H "Authorization: Bearer $APIMART_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mai-image-2.6",
    "prompt": "A travel poster of a university campus at dusk with the headline \"OPEN DAY 2026\", photorealistic, cinematic light",
    "size": "16:9",
    "resolution": "2K"
  }'

import os
import time
import requests

API_BASE = "https://api.apimart.ai"
headers = {"Authorization": f"Bearer {os.environ['APIMART_API_KEY']}"}

# 1. Submit the generation task
payload = {
  "model": "mai-image-2.6",
  "prompt": "A travel poster of a university campus at dusk with the headline \"OPEN DAY 2026\", photorealistic, cinematic light",
  "size": "16:9",
  "resolution": "2K"
}
resp = requests.post(f"{API_BASE}/v1/images/generations", headers=headers, json=payload, timeout=60)
resp.raise_for_status()
task_id = resp.json()["data"][0]["task_id"]

# 2. Poll the task until it finishes
for _ in range(200):
    task_resp = requests.get(f"{API_BASE}/v1/tasks/{task_id}", headers=headers, timeout=30)
    task_resp.raise_for_status()
    task = task_resp.json()["data"]
    if task["status"] == "completed":
        print(task["result"]["images"][0]["url"][0])
        break
    if task["status"] == "failed":
        raise RuntimeError(task.get("error", {}).get("message", "generation failed"))
    time.sleep(3)
else:
    raise TimeoutError(f"task {task_id} did not finish in time")

curl -X POST https://api.apimart.ai/v1/images/generations \
  -H "Authorization: Bearer $APIMART_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mai-image-2.6",
    "prompt": "Make the bicycle blue and add a small dog next to it",
    "image_urls": [
      "https://example.com/bicycle.png"
    ]
  }'

TASK_ID="task_xxx"  # replace with the task_id returned by the POST request
curl "https://api.apimart.ai/v1/tasks/$TASK_ID" \
  -H "Authorization: Bearer $APIMART_API_KEY"

MAI-Image-2.6 API Parameters

Parameter	Type	Required	Default	Description
`model`	string	Required	Default: mai-image-2.6	Model ID: mai-image-2.6, or mai-image-2.6-flash for the faster, lower-cost version.
`prompt`	string	Required	Default: —	Image description or editing instruction in English or Chinese, up to about 32,000 tokens.
`size`	string	Optional	Default: 1:1	Any integer aspect ratio from 1:4 to 4:1 (such as 16:9), a pixel size such as 1536x1024, or auto. Text-to-image only.
`resolution`	string	Optional	Default: 1K	1K or 2K, used with an aspect-ratio size. 4K is not supported and returns HTTP 400.
`width / height`	integer	Optional	Default: —	Exact pixels, sent as a pair. Each side at least 768, at most 2,359,296 total pixels; values round down to multiples of 32. Overrides size and resolution.
`image_urls`	array	Optional	Default: —	Up to 5 public image URLs or Base64 data URLs. One image edits it; several are combined. JPEG and PNG are supported; WebP and GIF are converted to PNG. Output dimensions are chosen by the model.
`auto_aspect_ratio`	boolean	Optional	Default: false	Set to true to let the model pick the aspect ratio from the prompt, the same as size: auto.
`web_grounding`	boolean	Optional	Default: false	Set to true to retrieve current information before generation, useful for real people, places or events.
`n`	integer	Optional	Default: 1	Only 1 is supported. Values above 1 return HTTP 400; submit separate tasks for more images.

MAI-Image-2.6 API FAQ

Model names, inputs, output limits, pricing and result handling

What is MAI-Image-2.6?

MAI-Image-2.6 is Microsoft AI's image generation and editing model, the quality tier of a family that also includes MAI-Image-2.6 Flash. It supports text-to-image, single-image editing and multi-image composition with up to five references. On APIMart, call it with the model ID mai-image-2.6.

When was MAI-Image-2.6 released?

Microsoft announced MAI-Image-2.6 on August 10, 2026. API access followed through Microsoft Foundry public preview, announced on September 4, 2026.

Is MAI-Image-2.6 Preview the same model as mai-image-2.6?

Yes. Microsoft Foundry lists the model as MAI-Image-2.6 (Preview) because it is in public preview there; Preview is a release status, not a separate model. The model is also written MAI Image 2.6. On APIMart, use the model ID mai-image-2.6.

What is the difference between MAI-Image-2.6 and MAI-Image-2.6 Flash?

Both models share the same capabilities and parameters. mai-image-2.6 is the higher-quality version for final assets; mai-image-2.6-flash is faster and costs less, with slightly lower image quality. Switch between them by changing the model field.

What parameters does the MAI-Image-2.6 API support?

model and prompt are required. Optional fields are size, resolution, width and height, image_urls (up to five references), auto_aspect_ratio, web_grounding and n, which must be 1. quality, style, background, output_format, response_format and mask_url are not supported and are ignored.

How many reference images can MAI-Image-2.6 edit with?

Provide up to five HTTP(S) image URLs or Base64 data URLs in image_urls. One image is edited; several are combined. JPEG and PNG are supported; WebP and GIF are converted to PNG. Public URLs must remain accessible.

What sizes can MAI-Image-2.6 text-to-image produce?

Use 1K or 2K with an integer aspect ratio from 1:4 to 4:1, or auto. Exact width and height must each be at least 768, with at most 2,359,296 total pixels. For example, 3072×768 is valid; 2048×2048 and 4K are not.

How do size settings interact with reference images?

For text-to-image, paired width/height override size and resolution; pixel-valued size overrides resolution. Exact dimensions round down to multiples of 32. With references, the model determines dimensions and these settings do not apply.

What does web_grounding do in MAI-Image-2.6?

Set web_grounding to true to use Bing retrieval before generation, especially for real people, places and events. It defaults to false. Retrieved context does not replace checking the final image.

Can I use masks or generate several formats at once?

No. Each request generates one PNG and n must be 1. Mask editing is not supported, and this page does not provide mask or output-format settings.

How much does the MAI-Image-2.6 API cost?

The MAI-Image-2.6 API charges actual text-input, reference-image and output-image tokens; see the pricing table on this page for current rates. Output-image tokens equal width × height ÷ 1024. An upfront charge is reconciled after completion. The displayed output estimate excludes text-input, reference-image and web-grounding costs, and failed tasks are fully refunded. Membership saves 20% (Gold) and up to 28% (Platinum and Diamond).

Is MAI-Image-2.6 free or open source?

No. APIMart has no free quota for MAI-Image-2.6; you pay per request from your balance. Microsoft offers the model as a hosted service and has not published downloadable weights, for example on Hugging Face, so it cannot be self-hosted.

How do I get a MAI-Image-2.6 API key?

Sign in to APIMart, add funds, then create a key on the API Keys page. The same key works for mai-image-2.6, mai-image-2.6-flash and every other model on the platform.

How do I call the MAI-Image-2.6 API and get the result?

Submit to POST /v1/images/generations, then query GET /v1/tasks/{task_id} every 3–5 seconds. Download completed images before expires_at (results are kept for 72 hours). A paused status check does not mean generation failed; resume querying the existing task. Full parameters are in the MAI-Image-2.6 docs at docs.apimart.ai.

Related Models

Explore more models in the same category.

View All Models →