Microsoft just made one of its boldest creative AI moves yet. At Build 2026 in San Francisco on June 2, the company’s AI Superintelligence team unveiled a family of seven homegrown models, and the one creators should care about most is MAI-Image-2.5, a brand-new text-to-image and image-to-image generator that is already climbing the public leaderboards. For anyone making art, marketing visuals, or social content, this is a genuinely free, genuinely capable new option that you can try today without a subscription.
Here at the Creative AI Show, we track every meaningful release in the text-to-image and image-to-video space, and MAI-Image-2.5 stands out because it is the first Microsoft model to handle both generating images from a prompt and editing images you already have. That dual capability, paired with free access, makes it one of the more interesting launches of the month.
What Is MAI-Image-2.5?
MAI-Image-2.5 is Microsoft’s latest in-house image model, built from scratch by the Microsoft AI team led by Mustafa Suleyman. It launched alongside a faster sibling, MAI-Image-2.5 Flash, which is tuned for high-volume, scalable production workloads where speed matters more than maximum fidelity. Microsoft describes the whole MAI lineup as a “family of models” built on a clean data lineage and engineered for efficiency, with the company claiming market-leading quality per dollar.
What sets this release apart from earlier Microsoft image models is its flexibility. MAI-Image-2.5 accepts image uploads alongside text prompts, so you can submit an existing photo with a modification instruction and get back an edited version. That puts it squarely in the same category as the image-editing tools from Google and OpenAI that creators have leaned on for the past year.
Text-to-Image and Image-to-Image in One Model
The headline feature is that MAI-Image-2.5 serves two creative workflows at once. On the text-to-image side, you type a prompt and the model renders an original image. On the image-to-image side, you upload a starting picture and describe the change you want, whether that is swapping a background, adjusting lighting, restyling a subject, or cleaning up a composition.
Microsoft is positioning the editing capability as especially valuable for real creative work. Rather than starting from a blank canvas every time, you can take a concept you already have and refine it iteratively. For designers, marketers, and content creators, this is often where the real value lives, since most projects involve enhancing existing assets rather than generating everything from zero.
How It Performs on the Arena Leaderboard
Benchmarks are not everything, but they are a useful sanity check, and MAI-Image-2.5 posted strong numbers on Arena, the community-voted leaderboard that pits image models against each other in blind comparisons. Here is where it landed:
- Text-to-image: #3 on Arena’s text-to-image leaderboard with a score of roughly 1,254, a 72-point jump over the previous MAI-Image-2, sitting behind only OpenAI’s gpt-image-2 and Google’s Nano Banana 2.
- Image editing: #2 on Arena’s single-image edit leaderboard with a score around 1,401, surpassing Google’s Nano Banana 2 and trailing only GPT-Image-2.
- The takeaway: a top-three finish in generation and a runner-up spot in editing puts Microsoft in elite company that, until recently, was held only by Google DeepMind and OpenAI.
For a model line that was barely in the conversation a year ago, breaking into the top tier on two separate leaderboards is a real statement. It signals that the image-generation race now has at least three serious frontier labs rather than two.
How to Try MAI-Image-2.5 for Free
The best part for creators on a budget is that you do not need to pay to experiment with MAI-Image-2.5. Microsoft has made it accessible through several free channels:
- Arena: the model is hosted on Arena for anyone to try in blind head-to-head comparisons, no paid subscription required.
- MAI Playground: Microsoft’s own sandbox for testing its MAI models directly.
- Microsoft Foundry: for developers who want to build the model into their own apps and workflows.
- PowerPoint and OneDrive: image generation is now live inside PowerPoint, with precise photo editing rolling out to OneDrive, so the tools meet you where you already work.
That spread of free entry points matters. Plenty of strong models launch behind paywalls or invite-only waitlists, so a frontier-grade generator you can poke at immediately is exactly the kind of release we love to highlight. If you only have a few minutes, start on Arena, where you can compare MAI-Image-2.5 side by side against the models it is chasing.
Why This Release Matters for Creators
Beyond the benchmarks, MAI-Image-2.5 is a sign of where the creative AI market is heading. Microsoft is deliberately weaning itself off its reliance on outside models and building a full multimodal stack of its own, spanning image, voice, transcription, reasoning, and code. For creators, more serious competition means faster improvements, better pricing, and more places to generate the visuals you need.
The Flash variant is worth a special mention for anyone producing content at scale. If you are batch-generating thumbnails, social posts, or product variations, a faster, cheaper model that still delivers strong quality can quietly transform your workflow. And because the same model family handles both generation and editing, you can stay in one ecosystem from first draft to final polish.
The Bottom Line
MAI-Image-2.5 is a confident, free-to-try entry into the crowded text-to-image and image-to-image arena, and its top-three leaderboard placements prove it is not just a vanity release. Whether you are a hobbyist exploring AI art or a professional refining client visuals, it is well worth a test drive on Arena or the MAI Playground this week.
We will keep tracking how MAI-Image-2.5 stacks up against the competition as more creators put it through its paces. For more hands-on breakdowns of the latest creative AI tools, visit the Creative AI Show, and subscribe to our YouTube channel for video walkthroughs and side-by-side tool comparisons.
No responses yet