Microsoft Launches MAI-Image-2: A Serious AI Image Generator Now Built Into Copilot
Microsoft has released its second-generation AI image model, ranking third globally and rolling out directly into Copilot and Bing Image Creator.
What happened
Microsoft has launched MAI-Image-2, its new AI image generation model, and it is a significant step forward from its predecessor.
The model now ranks third on the Arena.ai leaderboard, the most widely used benchmark for comparing text-to-image models, putting it behind only Google and OpenAI. MAI-Image-1 launched at ninth place, so this is a notable jump.
Key improvements include enhanced photorealism, with more accurate skin tones and natural lighting, and reliable in-image text generation, which has historically been a weak point for AI image tools. MAI-Image-2 can now generate posters, infographics, slides, and diagrams with readable text, something that has been difficult for competitors to do consistently.
The model is available now in the MAI Playground for experimentation, and is beginning to roll out across Microsoft Copilot and Bing Image Creator. API access is live today for select enterprise customers, with broader developer access to follow.
What this means for your business
If your business uses Microsoft 365 Copilot, you will likely have access to a noticeably better image generator in the coming weeks without doing anything.
For marketing, social media, and content teams, this matters. Creating product visuals, social posts, infographics, and presentation slides using AI has previously meant switching between tools. Having a competitive image generator built directly into Copilot, alongside the writing and productivity features your team already uses, simplifies that workflow.
The readable text capability is particularly useful for businesses. Generating an infographic or a branded poster with accurate text has been unreliable with most AI tools. If MAI-Image-2 delivers on this consistently, it removes one of the main reasons people still turn to designers for routine content tasks.
One thing to note: in its current playground form, there is a 24-hour limit after 15 images and output is limited to square format only. These are likely temporary constraints for the preview period, but worth knowing before building any production workflow around it.
Explore more on AdaHQ
Everything you need to start using AI in your business.