MAI-Image-2.7 and MAI-Image-2.7-Flash now available on AI Gateway
- Source
- vercel.com
- Author
- Kevin Dawkins
- Date

Context
Microsoft's MAI-Image-2.7 and MAI-Image-2.7-Flash are now in public preview on Vercel's AI Gateway, and the changelog names two practical changes over the previous generation. The largest output is now 2048 by 2048 pixels, up from the 1536 by 1536 that MAI-Image-2.6 offered on the gateway, and both models can produce transparent backgrounds for new and edited images. A developer asks for a transparent PNG by setting one option, so a product shot comes out as a cut-out object ready to place in a layout. Both models generate from text or edit an existing image from up to five reference images, through one gateway API key without a Microsoft account.
The split between the two follows the earlier pattern: MAI-Image-2.7 for final assets where quality comes first, Flash for faster, cheaper exploration and variations. Independent testing gives that choice some numbers. Artificial Analysis, which benchmarks AI models, ranks MAI-Image-2.7 first on its image-editing leaderboard at $34.82 per 1,000 images, a sixth of the price of the four models behind it, with Flash at $10.24 per 1,000.
- Vercel positions the two models as a quality versus cost split: MAI-Image-2.7 for final assets where image quality matters most, and the Flash variant for faster, cheaper runs while exploring prompts or making variations.
- Transparent output is not a separate model or endpoint: with AI SDK 6 or later you request it by setting providerOptions.azure.background to 'transparent' in the generateImage call.
- Editing works by passing existing images alongside a text instruction in the prompt object, and the models accept up to five reference images to guide an edit or combine subjects; the AI CLI exposes the same flow with an -i input flag.
Checking sign-in…
Loading comments…





