MAI-Image-2.5 Pro
MAI-Image-2.5 Pro is the quality-focused version of Microsoft AI's MAI-Image-2.5, released in public preview on 23 July 2026. Microsoft's model card describes a 20 billion parameter diffusion model, trained with flow matching and rebuilt for multimodal editing: object removal and replacement, attribute changes, inpainting, text updates and cleanup, with consistency across repeated edits. Output is capped at 1,048,576 pixels, the area of a 1024 square, in any shape. It is available through Microsoft Foundry and the MAI Playground, and on fal at token prices that come to about $0.17 an image.
Anvisha Pai, Co-founder & CEO, Voyager
Microsoft · released
›What we asked for
A self-portrait of MAI-Image-2.5 Pro, Microsoft's quality-focused image model: a gallery wall with one large framed photograph of a mountain lake at dawn, a small engraved brass plaque beneath it reading 'MAI-IMAGE-2.5 PRO', a visitor's silhouette in the foreground, soft museum lighting, photorealistic.
microsoft/mai-image-2.5-pro
aspect_ratio: 4:3, output_format: png, num_images: 1
returned 1024 by 768 pixels
At a glance
- Price
- About $0.17 a text-to-image request on fal, and $0.18 to $0.27 for an edit with one input image, from image tokens at $162 per million in and out.
- Sizes
- At most 1,048,576 pixels in any shape, in seven ratios from 16:9 to 9:16: 1024 square, 768 by 1024 at 3:4 and 1360 by 768 at 16:9 in our run. No 21:9.
- Edits
- Built for them, per Microsoft: removal, replacement, inpainting and text updates that keep the layout, over repeated passes. One input image per request.
- Speed
- 25 to 40 seconds a picture in our run, median about 29.
Other MAI-Image models: MAI-Image-2.5. Compare all 2.
Anvisha's take
Pro fixes most of what bothered me about the standard model, at three and a half times the price. The infographic's title is spelled right this time, the four steps exact on a photographed card; the chart has a clean 0 to 300 axis, every value printed right and April below March, though January's bar stands at 100, not 120; the composition test placed everything with three ice cubes; and the edit added a slim brass handle on the right instead of a ring on the rim, leaving the filter and grounds exactly as they were. The woodblock print signed itself with a legible Japanese title. It still sets the poster's date and ticket lines in capitals, it put a paper filter and coffee grounds in the product shot nobody asked for, and the output is small, 768 by 1024 at 3:4, with no 21:9. Worth it for edits you will ship; for drafts the standard model is enough.
MAI-Image-2.5 Pro real output: the same 12 prompts every model gets
Every image model on this site gets the same 12 prompts, so you can compare like with like. These are the pictures MAI-Image-2.5 Pro returned, one try each, nothing re-rolled or retouched. Generated on . The 12 pictures took 25 to 40 seconds each and cost $2.14 in total.
Spelling, kerning and hierarchy of three lines of set text.
›What we asked for
A concert poster that reads 'MODA SUMMER SESSIONS' in large bold type, with the subtitle 'Friday 12 June, 8pm, Roundhouse' and a smaller line 'Tickets at the door'. Risograph style, two colours, tangerine and navy, grainy ink texture.
microsoft/mai-image-2.5-pro
aspect_ratio: 3:4, output_format: png, num_images: 1
returned 768 by 1024 pixels
Catalogue lighting, a clean seamless background and a believable contact shadow.
›What we asked for
Studio product photograph of a matte black ceramic pour-over coffee dripper on a seamless white background, soft top light, subtle contact shadow, centred, catalogue style.
microsoft/mai-image-2.5-pro
aspect_ratio: 1:1, output_format: png, num_images: 1
returned 1024 by 1024 pixels
A natural face, correct finger count and hands holding an object.
›What we asked for
Close-up portrait of a woman in her 40s with short grey hair, laughing, holding a red ceramic mug with both hands, natural window light, 85mm lens, shallow depth of field.
microsoft/mai-image-2.5-pro
aspect_ratio: 3:4, output_format: png, num_images: 1
returned 768 by 1024 pixels
A flat single-colour mark with a legible wordmark and no gradients or photo texture.
›What we asked for
A flat vector logo mark for a bicycle repair shop called 'Spoke & Chain': a single-colour icon combining a chain link and a wheel spoke, with the wordmark below in a geometric sans-serif, black on white.
microsoft/mai-image-2.5-pro
aspect_ratio: 1:1, output_format: png, num_images: 1
returned 1024 by 1024 pixels
Many distinct objects and people in one frame without merging or duplicating them.
›What we asked for
A crowded Saturday farmers market seen from slightly above: at least twelve stalls with striped awnings, shoppers carrying tote bags, a dog on a leash, crates of oranges, a musician with a guitar, bunting overhead, late morning light.
microsoft/mai-image-2.5-pro
aspect_ratio: 4:3, output_format: png, num_images: 1
returned 1024 by 768 pixels
A named historical style applied faithfully rather than a generic painterly look.
›What we asked for
A lighthouse on a rocky cliff during a storm, in the style of a Japanese ukiyo-e woodblock print, with visible wood grain texture, flat colour areas and a bokashi gradient sky.
microsoft/mai-image-2.5-pro
aspect_ratio: 3:4, output_format: png, num_images: 1
returned 768 by 1024 pixels
Six labelled bars at the stated heights, in order, with readable axes.
›What we asked for
A clean bar chart titled 'Monthly active users' showing six months, January through June, with the values 120, 150, 180, 170, 220 and 260, labelled axes, white background, one accent colour, no 3D effects.
microsoft/mai-image-2.5-pro
aspect_ratio: 4:3, output_format: png, num_images: 1
returned 1024 by 768 pixels
A 21:9 composition that leaves the left third clear for headline text.
›What we asked for
Website hero banner: a wide panorama of a sunlit modern co-working space with plants, the left third of the frame kept as clear wall space for text, warm neutral palette.
microsoft/mai-image-2.5-pro
aspect_ratio: 16:9, output_format: png, num_images: 1
returned 1360 by 768 pixels
Four numbered steps in the given order with their exact wording, plus an icon each.
›What we asked for
An infographic card titled 'How to brew pour-over' with four numbered steps in this order: 1. Rinse the filter, 2. Bloom for 30 seconds, 3. Pour in slow circles, 4. Wait 3 minutes. A simple icon beside each step, muted palette, clean sans-serif type.
microsoft/mai-image-2.5-pro
aspect_ratio: 3:4, output_format: png, num_images: 1
returned 768 by 1024 pixels
Appetising, menu-grade food with correct textures and no melted anatomy.
›What we asked for
Overhead photograph of a margherita pizza on a wooden board, fresh basil leaves, one slice pulled out with the cheese stretching, restaurant menu photography, natural light.
microsoft/mai-image-2.5-pro
aspect_ratio: 1:1, output_format: png, num_images: 1
returned 1024 by 1024 pixels
Exact counts and left-right placement of five named objects.
›What we asked for
A glass of water on a wooden table with exactly three ice cubes in it, a yellow pencil resting across the rim of the glass, a blue book to the left of the glass and a red apple to the right, photographed from table height.
microsoft/mai-image-2.5-pro
aspect_ratio: 4:3, output_format: png, num_images: 1
returned 1024 by 768 pixels
Change one object's colour and add a part while leaving background, light and framing untouched.
›What we asked for
Change the coffee dripper to a deep forest green and add a small brass handle on its right side. Keep the background, lighting and composition exactly as they are. (applied to the product on white result above)
microsoft/mai-image-2.5-pro/edit
aspect_ratio: auto, output_format: png, num_images: 1
returned 1024 by 1024 pixels
How the cost under each picture was worked out: fal's average ~$0.17 per text-to-image request for MAI-Image-2.5 Pro (image output $162 per million tokens, text input $7.50 per million) (fal.ai/models/microsoft/mai-image-2.5-pro, read 2026-10-22); fal does not return the charge.
Compare MAI-Image-2.5 Pro
MAI-Image-2.5 Pro pricing
| Where | Unit | Price | Source |
|---|---|---|---|
| fal | Image, average text-to-image request Image output at $162 per million tokens, text input at $7.50 per million. | $0.17 | fal model page, read |
| fal | Edit with one input image, top of fal's estimate fal estimates $0.18 to $0.27 per output image; image input and output are both $162 per million tokens. | $0.27 | fal model page, edit, read |
Is MAI-Image-2.5 Pro free?
No free API tier. Microsoft offers Pro in its MAI Playground and through Foundry under those services' terms; on fal it is billed by tokens.
- •MAI Playground and Microsoft Foundry: Microsoft's own routes, in public preview from 23 July 2026.
- •fal: about $0.17 an image.
Where to use MAI-Image-2.5 Pro
- Microsoft Foundry · Microsoft's developer catalogue.
- fal · microsoft/mai-image-2.5-pro and /edit, token-billed; the endpoints used for the samples below.
- Voyager · Runs inside Voyager, the agent for creative work, with your own fal key.
Voyager
Run MAI-Image-2.5 Pro inside Voyager
Voyager is an agent for creative work that runs every model in this reference, MAI-Image-2.5 Pro included. Brief it, and it picks the model, makes the thing and hands back something you can edit. Create an account to get started.
What we noticed running MAI-Image-2.5 Pro
- •Forest green with a slim brass handle on the right; the filter paper, grounds, seam and background are unchanged. See the edit an existing image picture.
- •A clean 0 to 300 axis and every value printed right, with April below March, but January's bar stops at 100, not 120. Months are abbreviated and it added '(thousands)'. See the chart picture.
- •The four steps and title exact, as a card photographed on a table beside a dripper, with an 'Enjoy your coffee' footer added. See the infographic with steps picture.
- •Book left, apple right, three ice cubes and the pencil across the rim; the book gained a 'NOTES' title. See the hard composition picture.
- •A seamed cone with a pouring lip, holding a paper filter full of coffee grounds that were not asked for; no handle or stand. See the product on white picture.
- •All three lines spelled right, the date and ticket lines in capitals, over a band of guitarist, singer and drummer. See the text in an image picture.
- •A strong print with a legible Japanese title cartouche, '嵐の灯台' (lighthouse in the storm), and a seal. See the specific art style picture.
- •16:9 at 1360 by 768, the widest ratio offered, with the left third a bare, sunlit wall. See the wide aspect ratio picture.
- •The self-portrait spelled 'MAI-IMAGE-2.5 PRO' correctly on the small brass plaque.
MAI-Image-2.5 Pro limits and API parameters
| Limit | Value |
|---|---|
| Output size | At most 1,048,576 pixels in total; either side can pass 1024 if the total stays within it. |
| Context | 32,000 tokens of input; prompts up to 5,000 characters on fal. |
| Aspect ratios | Seven, from 16:9 to 9:16. fal API schema, read |
›Every setting, for developers
fal (microsoft/mai-image-2.5-pro and /edit)
fal API schema, read
| Parameter | Type | Values | Default |
|---|---|---|---|
| prompt | string | Up to 5,000 characters Required. On the edit endpoint, the instruction. | |
| aspect_ratio | enum | auto, 1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3 | auto |
| image_url | string | One URL Edit endpoint only. Unlike the standard model's edit, a single string rather than a list. | |
| num_images | integer | 1 to 4 | 1 |
| output_format | enum | jpeg, png, webp | png |
What MAI-Image-2.5 Pro will not draw
Microsoft says it aligned the model to reduce harmful or inappropriate images, such as violent, gory or sexual content, even when requested, with mitigations in the training data and at deployment. The licence and acceptable use are those of the service it runs through, such as Azure AI Foundry or MAI Playground. fal marks it for commercial use and exposes no safety setting.
Alternatives to MAI-Image-2.5 Pro
Frequently asked questions
What is MAI-Image-2.5 Pro?
Microsoft AI's quality-focused image model, a 20 billion parameter version of MAI-Image-2.5 rebuilt for precise editing, released in public preview on 23 July 2026.
How much does MAI-Image-2.5 Pro cost?
About $0.17 a text-to-image request on fal, and $0.18 to $0.27 for an edit with one input image, billed by tokens, read on 22 October 2026.
What resolution does it output?
At most about one megapixel, 1,048,576 pixels, in any of seven ratios. Our 3:4 pictures were 768 by 1024.
Is Pro worth it over MAI-Image-2.5?
On our prompts it fixed the standard model's misspelled infographic and ring-shaped edit handle and placed the composition test correctly. It costs about three and a half times as much.
How this page is made
Prices, limits and settings are read from the linked vendor and host pages on the dates shown. Every picture was generated by us through fal with the request shown under it, and the original files are kept unedited. A generation that failed is shown as a failure. Model and vendor names are used to identify the products; no vendor imagery appears on this page.
