Microsoft unveils its most powerful AI-powered raw image model, MAI-Image-2.5.

Microsoft Research has launched MAI-Image-2.5, the most powerful image generation model, which has risen to 3rd place on the Arena text-to-image leaderboard, breaking the previous dominance of Google DeepMind and OpenAI in the top 5. This represents a 72-point improvement over its predecessor. The model significantly enhances text rendering capabilities, making it suitable for generating infographics, posters, and other commercial materials. It also shows significant improvements in visual reasoning, image detail, and stylistic consistency.