Microsoft Reveals its First Image Generation AI Model

Microsoft announced yesterday MAI-Image-1, its first image generation model designed to create photorealistic imagery. It follows the release of MAI-Voice-1 and MAI-1-Preview in August, which were the company’s first two in-house AI models.

“MAI-Image-1 marks the next step on our journey and paves the way for more immersive, creative and dynamic experiences inside our products. We trained this model with the goal of delivering genuine value for creators, and we put a lot of care into avoiding repetitive or generically-stylized outputs,” the company said yesterday.

While OpenAI remains Microsoft’s main partner on “frontier” AI models, the Redmond giant appears to be doubling down on in-house AI models. In an interview with The Verge back in September, Microsoft AI CEO Mustafa Suleyman said that the company “should have the capacity to build world class frontier models in house of all sizes.”

At the moment, Microsoft is testing MAI-Image-1 in LMArena, an open platform for evaluating AI models. As of this writing, MAI-Image-1 ranks #9 on the platform’s Text-to-Image leaderboard, while Google’s Nano Banana model is a close second to the Hunyuan Image 3.0 model from Tencent.

Microsoft says that it developed and trained MAI-Image-1 while taking into account feedback from professionals in the creative industries. The company plans to make MAI-Image-1 available in Copilot and Bing Image Creator soon as an alternative to OpenAI’s models.