Microsoft has released MAI-Image-2, its second-generation in-house text-to-image AI model, which has ranked #3 on the Arena.ai leaderboard for image generation. The model places directly behind competitors from Google and OpenAI, marking a significant improvement from its predecessor, MAI-Image-1, which debuted at the #9 position.
The new model, developed by Microsoft's AI Superintelligence team, focuses on enhanced photorealism, including more accurate lighting and skin tones, and improved in-image text generation for tasks like creating diagrams and infographics. MAI-Image-2 is being rolled out in the MAI Playground for public testing and will be integrated into products like Copilot and Bing Image Creator, with API access available for select customers.