On May 26, 2026, Microsoft Research introduced its newest image generation model, MAI-Image-2.5, which swiftly ascended to the third position on the Arena text-to-image model leaderboard. Touted as being 'closer to commercial readiness,' this model marks a significant leap in text rendering precision over its forerunner, MAI-Image-2. This enhancement renders it particularly apt for applications demanding exact text depiction, like infographics and promotional posters. Moreover, it bolsters its proficiency in crafting stylized illustrations and commercial imagery, offering a superior level of visual polish. The model also refines its visual reasoning skills, gaining a deeper grasp of object compositions, lighting effects, and spatial dynamics. Presently accessible on the Arena platform, the model is slated for integration into MAI Playground and Foundry within the forthcoming two weeks.