Microsoft Launches Its First AI Image Model, MAI-Image-1
Now available in Bing Image Creator and Copilot, the model enters a crowded field with a top-10 ranking and a focus on visual diversity
- MAI-Image-1 debuts in top 10 on LMArena
- Available in Bing Image Creator and Copilot Audio Expressions
- Excludes EU countries for current access
Microsoft has released MAI-Image-1, its first in-house AI model for image generation, now live in Bing Image Creator and select Copilot features. As of November 4, the model is officially ranked in the top 10 on LMArena, a benchmarking site for text-to-image models.
Users can access MAI-Image-1 through bing.com/create, the Bing mobile app, or directly via the Bing search bar. It appears as an option alongside other models like DALL·E 3 and GPT-4o, giving users the ability to choose between different styles and outputs depending on their creative needs.
Built for stories and search
The model is also integrated into Copilot Audio Expressions, a feature that converts spoken stories into AI-generated visuals. In Story Mode, MAI-Image-1 handles image generation based on the user’s narration or prompts.
While Microsoft hasn’t disclosed much about the training data or architecture, the company noted that the model was developed with “rigorous data selection” and tested to avoid generic or repetitive results. The aim is to produce more flexible and diverse visual output—something increasingly difficult as generative image tools begin to converge in look and feel.
Regional limits and rollout
MAI-Image-1 is currently available in most regions where Bing Image Creator and Copilot Labs are supported, but not in the European Union. Microsoft hasn’t provided a reason for the regional exclusion, though regulatory friction around generative AI tools in the EU has delayed other rollouts before.
A step into crowded territory
By entering the image generation space with its own model, Microsoft is moving closer to owning more of its AI stack. Until now, the company has leaned on OpenAI’s models for generative tasks. MAI-Image-1 marks a shift toward more internally developed tools, particularly for consumer-facing applications.
Related reading
- Claude Opus 4.5 Lands on Microsoft Foundry, and It Is Aiming Straight at Enterprise Workflows
- Foxconn Just Landed a Major OpenAI Deal. Is This the Start of an AI Manufacturing Power Shift?
- Copilot introduces holiday mode
There’s no claim here of being the best or fastest. Instead, MAI-Image-1 is being positioned as a solid, flexible alternative—built to slot neatly into existing tools without overwhelming users or developers with new systems to learn.
In a field dominated by loud launches and hyperbolic demos, Microsoft’s approach with MAI-Image-1 is notably muted. It’s a model that works in the background, fits into workflows, and avoids the flash of overpromising. For now, that may be exactly the point.
