Microsoft Releases MAI-Image-2.7-Preview, Ranking Second Globally in Image Editing
Microsoft has released the MAI-Image-2.7-Preview model, which currently ranks as the second-best AI image editing model globally on the Artificial Analysis benchmark. Announced by Microsoft AI CEO Mustafa Suleyman, the model achieves a score of 1,193 Elo, outperforming competitors from Google, Meta, and xAI. This release demonstrates Microsoft's strengthening position in the competitive generative AI and computer vision landscape, directly challenging market leaders. Its high ranking in complex tasks like style transfer and text modification highlights significant advancements in controllable image manipulation. The preview model focuses exclusively on image editing using original images and text prompts, while text-to-image generation is slated for a later release. Currently, it is available via MAI Playground, Arena, and Artificial Analysis, though public APIs, model weights, and commercial pricing have not yet been announced.
## BACKGROUND
Artificial Analysis provides independent evaluations and benchmarks for various artificial intelligence models, including text and image generators, using blind voting and Elo ratings to measure performance. Generative AI and computer vision technologies enable machines to understand, interpret, and manipulate visual data based on human instructions.