Microsoft MAI-Image-2.5 for photorealistic and artistic text-to-image generation.
Added Jun 3, 2026
Approx. Price
$0.005 per image
Model Type
text-to-image
Settings
Generation controls available for this model.
Images Per Run
Up to 1
Output images
Input Images
Up to 1
Reference/edit images accepted • Route max 30 MB
Output Sizes
Aspect Ratio
Default
auto
Number of Images
Default
1
Resolution
Default
auto
Options (8)
auto (Auto), 1:1 (Square), 16:9 (Landscape), 9:16 (Portrait) +4 more
Benchmarks
Benchmarks
Human preference benchmarks sourced from Artificial Analysis.
Text to Image
#8 / 159
ELO
1105.0
Appearances
15,220
95% CI
-9/9
Image Editing
#7 / 76
ELO
1108.0
Appearances
12,154
95% CI
-9/9
Release Date 2026-06 · Matched as MAI-Image-2.5
Artificial Analysis APIExamples
Loading examples…
Related image models
Compare MAI-Image-2.5 with similar models from the same provider or model family.
MAI-Image-2.6
microsoft/mai-image-2.6Microsoft’s image model for detailed photography, product imagery, accurate lettering, and precise edits to an uploaded image.
MAI-Image-2.6 Flash
microsoft/mai-image-2.6-flashA faster, lower-cost MAI image model for product visuals, creative drafts, and edits to an uploaded image.
BiRefNet V2
fal-ai/birefnet/v2Remove backgrounds with dedicated foreground segmentation, preserving transparent edges. Optionally return the mask itself.
GPT Image 2.5 Flare Edit
openai/gpt-image-2.5/flare/editGPT Image 2.5 Flare edits images from natural-language instructions with up to 16 reference images. Supports quality levels from low to max and 1K, 2K, or 4K output. Pricing depends on quality, resolution, and reference image count.
GPT Image 2.5 Flare
openai/gpt-image-2.5/flare/text-to-imageGPT Image 2.5 Flare generates images from text prompts or edits images with up to 16 reference images. Supports quality levels from low to max and 1K, 2K, or 4K output. Pricing depends on quality, resolution, and reference image count.
GPT Image 2.5 Sunburst Edit
openai/gpt-image-2.5/sunburst/editGPT Image 2.5 Sunburst edits images from natural-language instructions with up to 16 reference images. Supports quality levels from low to max and 1K, 2K, or 4K output. Pricing depends on quality, resolution, and reference image count.