Microsoft MAI-Image-2

Microsoft MAI-Image-2.7 debuts, claiming to be "the world's second image editing model." Microsoft released MAI-Image-2.7-Preview on October 27. AI CEO Suleiman declared: MAI-Image has risen to second place in the world in the Artificial Analysis image editing list, second only to OpenAI's GPT Image 2.5 Sunburst (max), surpassing Google, Meta, xAI and OpenAI Another GPT-Image-2.5-Flare is left behind. It is not a Vincent picture, but a controllable editor - feed an original picture and add text commands, and output the modified picture. It is good at the "difficult to correct" tasks: changing the overall style, re-lighting, modifying text symbols on the screen, repairing and enhancing materials, and complex changes that require an understanding of the relationship between objects, light and environment. Four capabilities are at the forefront. Compared with 2.6, 2.7 has improved in all 10 use cases and 7 types of editing actions of Artificial Analysis, and the four categories of enhanced repair, inferential editing, scene style, and text symbols have reached the forefront of the category. Scene and style have the largest lead, 24 Elo higher than the next best model; on the AA-Image-Editing v2.1 list, it ranks second with 1193 Elo. Users can experience it in Arena, Artificial Analysis and MAI Playground. Currently, only image editing is available. Vincent Picture is only said to be "coming soon" without a date. As a Preview, it has not yet announced its public API, pricing, and commercial availability. Model weights are not yet open, and it cannot be self-hosted. Availability in the Playground region may also be limited. If you want to shake OpenAI's top spot, you have to wait for Wen Shengtu to complete it. via AI News (author: AI Base)