Microsoft Unveils MAI-Image-1, Its First In-House Image Generation Model

Microsoft AI has announced MAI-Image-1, its first fully in-house image generation model, now ranked among the top 10 text-to-image models on LMArena.

Following its first in-house language models introduced in August, Microsoft says MAI-Image-1 is built to power more immersive, creative, and dynamic experiences across its ecosystem, including Copilot and Bing Image Creator.

Trained with a focus on artistic quality and realism, MAI-Image-1 aims to avoid repetitive or overly stylized results. The company emphasised rigorous data selection and feedback from creative professionals to ensure the model reflects authentic visual diversity and supports professional workflows.

The model demonstrates particular strength in producing photorealistic imagery, including complex lighting, landscapes, and textures—achieving high visual fidelity at greater speed than many larger models. This balance of speed and quality enables creators to iterate ideas quickly and refine their work seamlessly across tools.

"MAI-Image-1 excels at generating photorealistic imagery, like lighting (e.g., bounce light, reflections), landscapes, and much more. This is particularly so when compared to many larger, slower models," Microsoft said in a blog post.

Microsoft has begun testing MAI-Image-1 on LMArena to gather insights and improve safety and reliability ahead of broader rollout. “We want to deliver genuine value for creators while ensuring responsible and safe outcomes,” the company said.

MAI-Image-1 will soon be available in Copilot and Bing Image Creator, extending Microsoft’s commitment to accessible and responsible AI innovation.

For those eager to shape the next generation of AI models, Microsoft AI is actively recruiting for roles in its research and product teams.