news
Microsoft calls MAI Image 2.7 the top image editor on Artificial Analysis, while Arena ranks it third and fourth
Microsoft says the model beats GPT Image 2.6 Sunburst at one fifth the price, Artificial Analysis says one sixth against the four models behind it, and Arena places it third on multi-image edit and fourth on single.
Microsoft's new image model collected a first-place claim and a third-and-fourth-place result on the same day, from two different evaluators, with nothing in the announcement explaining the gap. MAI Image 2.7 and MAI Image 2.7 Flash went out at Ignite on November 17, 2026, alongside MAI Code 2, MAI Code 2 Flash and MAI Cyber 2 Flash, with Microsoft framing the image models around editing quality and price per thousand images.
First on one editing board, third and fourth on the other
Microsoft's headline claim is that MAI Image 2.7 ranks first on Artificial Analysis for image editing, ahead of GPT Image 2.6 Sunburst (max), at one fifth of the price. Artificial Analysis's own launch-day post calls it the new number one on AA-Image-Editing v2.1 and puts MAI Image 2.7 at $34.82 per thousand images and MAI Image 2.7 Flash at $10.24, with both landing on its quality versus price frontier for image editing and text to image.
Arena, running a separate vote-based evaluation, published a different placement on the same day: third on multi-image edit, fourth on single-image edit, and Microsoft AI as the second-ranked lab across all three of its image arenas, at $0.035 per image. The two price figures are consistent with each other, since $0.035 per image is $35 per thousand. The rankings are not. A model can be first on a pairwise board with a judged editing rubric and fourth on a crowd-voted one, and that difference is the whole argument about which board to quote, but neither Microsoft's announcement nor the surrounding coverage reconciles the two.
The preview build of this model was already close to the top of the Artificial Analysis editing board, sitting second by six Elo earlier in November, so the move to first is a small shift in a tight band rather than a jump.
One fifth or one sixth, depending on the comparison set
The price-advantage claim also exists in two forms. Microsoft says one fifth the price of GPT Image 2.6 Sunburst (max), a single named competitor. Artificial Analysis says one sixth the price of the four models ranked behind it, an average across a group. Both can be true at once, because they are different calculations, but a reader who sees only the headline number will not know which comparison produced it.
Per-unit prices quoted at one endpoint also tend not to survive contact with the resale market. On our price radar, a single tracked model carries different rates on every platform that hosts it, which is why a leaderboard's cost axis describes one listed endpoint rather than what any given buyer pays. Microsoft lists MAI Image 2.7 through Foundry and through partners including fal, Genspark, OpenRouter and Vercel.
The access status is written two ways
Microsoft's Ignite Book of News says MAI-Image-2.7 and MAI-Image-2.7-Flash are in preview and will offer sharper editing and more consistent image generation. Arena's post describes the model as out of preview. Trade coverage splits the difference, reporting public preview on Foundry and through partners, with Flash powering a Canvas workspace in MAI Labs. For anyone trying to work out whether the ranked model is the one they can deploy, that distinction is the practical question, and it currently has two answers.
One more item from the same announcement touches evaluation directly: Microsoft-Decision-1.1 is listed as generally available with what Microsoft describes as efficient multimodal routing and verification. A verification model that takes multimodal input is, in effect, a judge, and the Book of News entry carries no accuracy figures for it.
Sources: Microsoft Ignite 2026 Book of News, TestingCatalog.
Cite this
Free to cite and reuse with a link back. Data is updated as new runs and prices come in, so include the date.
Genativity. (2026). Microsoft calls MAI Image 2.7 the top image editor on Artificial Analysis, while Arena ranks it third and fourth. Retrieved November 18, 2026, from https://genativity.com/news/mai-image-2-7-first-on-aa-third-and-fourth-on-arena<a href="https://genativity.com/news/mai-image-2-7-first-on-aa-third-and-fourth-on-arena">Microsoft calls MAI Image 2.7 the top image editor on Artificial Analysis, while Arena ranks it third and fourth</a> (Genativity)Daniel Ochoa: Covers model launches, shutdowns and pricing changes as they happen. Reads deprecation notices for a living so you do not have to.