Microsoft releases MAI-Image-2.7-Preview for advanced image editing

Microsoft has announced the launch of MAI-Image-2.7-Preview, a new AI model specifically designed for advanced image editing and manipulation. According to Mustafa Suleyman, CEO of Microsoft AI, the model has achieved the second-place ranking globally on the Artificial Analysis image editing leaderboard, trailing only OpenAI’s GPT Image 2.5 Sunburst (max).

Unlike many generative models that focus on text-to-image creation, MAI-Image-2.7-Preview prioritizes the modification of existing images. The model is designed to handle complex tasks such as changing scene lighting while maintaining stylistic consistency, moving objects within an environment, enhancing image quality, and modifying text or symbols within a frame while preserving natural integration.

Microsoft reports that the model shows improvements across ten use cases and seven types of editing actions compared to its predecessor. While text-to-image capabilities are expected to be released in the future, the current preview version is focused on controllable editing for photos, designs, and media assets. The model is currently available for testing via Arena, Artificial Analysis, and the MAI Playground.