Microsoft
MAI-Image 2.6
Best for photoreal people, complex scenes and accurate text.
What MAI-Image 2.6 is good at
MAI-Image 2.6 is Microsoft’s newest image model and ranks fourth on the Arena text-to-image leaderboard. It is strongest on photoreal people and scenes with many elements.
Strong at
- Photoreal people with natural skin texture
- Complex scenes with several subjects
- Accurate text inside the image
- Edits that combine up to five reference photos
Keep in mind
- Slower than most image models, around 30 seconds per image
- PNG output only
In Phosphene
- 4 credits per image
MAI-Image-2.6
Microsoft’s newest image model via OpenRouter. Photoreal people, complex compositions, accurate text, and edits from up to five references.
Prices are per image at 1K and standard quality. Higher resolution costs more; the composer shows the exact price before you generate.
Questions about MAI-Image 2.6
- How much does MAI-Image 2.6 cost in Phosphene?
- 4 credits per image at 1K and standard quality. Higher resolutions cost more.
- Who makes MAI-Image 2.6?
- Microsoft AI. Phosphene runs it through OpenRouter, so you do not need a Microsoft or OpenRouter account.
- How many reference photos can MAI-Image 2.6 use?
- Up to five. Attach them in the composer and say what each one is for.
- Why does it take longer than other models?
- MAI-Image 2.6 spends more time per image, about 30 seconds. If you need speed, Nano Banana 2 or GPT Image 2.5 Flare return in under 10 seconds.
Other image models
- GPT Image 2.5Prompt adherence, readable text, product and brand images
- Nano Banana 2High-quality images fast, clear text and reference edits
- Seedream 5Campaign images, multilingual typography and realism
- FLUX 3Text-rich layouts, photoreal scenes and edits from many references