Microsoft's superintelligence team ships MAI-Image-2, a text-to-image generator

Microsoft's new superintelligence team has unveiled its first product: MAI-Image-2, an image generator rolling out across Microsoft's own products and eventually available via API.

Microsoft's superintelligence team, led by Mustafa Suleyman, has released MAI-Image-2, an AI model that turns text prompts into images. The model currently ranks third on the Arena.ai leaderboard for text-to-image generators, trailing OpenAI's GPT-Image-1.5 and Google's Nano Banana 2 by a significant margin.

According to Microsoft, MAI-Image-2 produces especially realistic photos with natural lighting and accurate skin tones, while also handling detailed and surreal scenes. The company says it built the model alongside photographers, designers, and visual artists.

The model also does well with more practical tasks, like reliably rendering text in images for posters, infographics, or diagrams.

MAI-Image-2 is available for testing in the MAI Playground (depending on region) and will roll out to Copilot and Bing Image Creator. API access is limited to select business customers for now but will open to all developers through Microsoft Foundry soon. Microsoft hasn't shared technical details, pricing, or info about training data.

Microsoft shipped its first in-house image generator, MAI-Image-1, in October 2025. It landed in ninth place on the Arena.ai leaderboard and didn't make much of a splash. MAI-Image-2's jump to third indicates a step forward, though Microsoft still has ground to cover before it can compete with the top models from OpenAI and Google, something the company reportedly plans to pursue.