Microsoft launches MAI-Image-1, its own image generation model
Microsoft continues its shift toward proprietary artificial intelligence. Its new image generation model, MAI-Image-1, is now available in Bing Image Creator and Copilot Audio Expressions, marking a significant step in the American giant’s strategy to reduce its reliance on OpenAI.
MAI-Image-1: A 100% Microsoft Model
Announced in October, MAI-Image-1 is the first visual creation model entirely designed by Microsoft’s internal AI teams. According to Mustafa Suleyman, director of Microsoft AI (and co-founder of DeepMind), the model will soon be deployed in Europe.
“MAI-Image-1 excels at creating images of food, natural landscapes, and artistic scenes with realistic lighting effects,” Suleyman stated on X.
On its official blog, Microsoft highlights the advantages of MAI-Image-1:
- Enhanced photorealism, particularly in handling light, reflections, and texture.
- Rapid execution, allowing for smoother generation than competing models of comparable size.
- Agile iteration, ideal for creators looking to quickly refine their ideas before exporting them to other tools (such as Paint, Designer, or Photoshop).
The company claims that MAI-Image-1 competes in quality with heavier models while being faster and more resource-efficient.
An Image AI Integrated with Copilot Audio Expressions
MAI-Image-1 is not limited to Bing. The model also generates automatic illustrations for audio stories created in Copilot Audio Expressions, a storytelling platform based on AI voice synthesis.
Specifically, when Copilot creates an audio story, MAI-Image-1 produces artistic visuals synchronized with the tone and theme of the narrative, offering an immersive experience that combines voice and image.
A Component of the New MAI Ecosystem
MAI-Image-1 joins the family of “MAI” (Microsoft AI) models, alongside:
- MAI-Voice-1, a natural speech synthesis vocal model,
- MAI-1-preview, a text model that serves as the foundation for certain functions of Copilot.
This diversification confirms Microsoft’s pivot toward a hybrid AI, combining its own models with those from partners like OpenAI and Anthropic.
Currently, Copilot utilizes GPT-5 (OpenAI) as its main model while allowing the activation of Claude (Anthropic) or the “MAI” models for specific tasks.
Three Models Available on Bing Image Creator
On the Bing Image Creator platform (web and app), users can now choose from:
AI Model | Origin | Specialty |
|---|---|---|
MAI-Image-1 | Microsoft | Speed, realism, landscapes, nature |
DALL-E 3 | OpenAI | Artistic and conceptual creations |
GPT-4o (visual) | OpenAI | Multimodal generation (text + image) |
This integration illustrates Microsoft’s intention to offer a unified creative experience, where users can choose the AI engine that best suits their needs.
A Strategic Alternative to OpenAI
Since launching the “MAI” range, Microsoft has begun a partial disengagement from its technological dependence on OpenAI. Its in-house models offer greater commercial and regulatory flexibility, especially in light of European compliance requirements.
The upcoming deployment of MAI-Image-1 in the EU is set to be the first large-scale test of a fully Microsoft-branded generative AI.
With MAI-Image-1, Microsoft takes a significant step toward creative autonomy. Faster than DALL-E 3 and more accessible than professional tools like Midjourney, the model embodies Suleyman’s vision — a useful, efficient AI integrated into daily productivity.
After words and voice, Microsoft aims to master images. MAI-Image-1 is just the beginning.
