Microsoft launches its second generation AI-photo maker, ranks third after…

Microsoft has launched MAI-Image-2, the second generation of its AI-powered image generation model to compete with Google and OpenAI. The model arrives with a notable milestone attached: It ranks third on the Arena.ai leaderboard, one of the most widely referenced benchmarks for comparing AI image generation tools. This also places Microsoft among the top text-to-image labs in the world.

The announcement marks a significant step up for Microsoft's MAI model family because its MAI-Image-1 made its debut at the 10th spot on the LMArena leaderboard, suggesting that Microsoft's AI image capabilities are maturing quickly. MAI-Image-2 is behind Google Gemini (gemini-3.1-flash-image-preview (nano-banana-2) and OpenAI (gpt-image-1.5-high-fidelity).

Microsoft says new text-to-image model built with creatives in mind

Microsoft says that MAI-Image-2 was built with a specific user in mind: the working creative. Microsoft says that before developing the model, Microsoft’s team spoke directly with photographers, designers and visual storytellers to understand where AI image generation was still falling short in everyday professional use.




“MAI-Image-2 is built for creatives who want images that feel like they exist in the world, with natural light, accurate skin tones, environments that feel lived-in. Creatives can now spend less time fixing in post-production and more time making,” Microsoft said. Three areas emerged as priorities: Photorealism, Text generation within images and complex and detailed scene generation – and MAI-Image-2 has been built around all three.

Photorealism is the first: The model is designed to generate images that feel grounded in the real world. For example, they exhibit accurate skin tones, natural lighting and environments that look lived-in rather than artificially constructed, Microsoft claims. The aim is to reduce the amount of time creatives spend correcting AI-generated images in post-production.

Text generation within images is the second: Getting AI to reliably render readable, well-placed text inside an image has been one of the weaknesses, and MAI-Image-2 addresses this directly, enabling consistent creation of infographics, slides, diagrams, posters and scenes where signage or typography plays a key role.

Complex and detailed scene generation is the third: For creatives working on cinematic compositions or creating elaborate fantasy worlds, Microsoft says, the MAI-Image-2 is designed to handle the creation of intricate designs without losing coherence or detail.

How users can test it

MAI-Image-2 is available to try right now in the MAI Playground, Microsoft's dedicated environment for experimenting with its latest AI models. Users can generate images and share feedback directly with the development team. Beyond the playground, the model is beginning to roll out across Copilot and Bing Image Creator.