Microsoft Opens MAI-Image-2.6 to Developers in Foundry, Adds a Faster Flash Version

Microsoft AI said on September 4, 2026 that it is bringing the MAI-Image-2.6 image model to developers in Microsoft Foundry and expanding the family with MAI-Image-2.6-Flash for latency-sensitive workloads. Both support multi-image reference editing, web grounding and dynamic aspect ratios.

Microsoft image-generation API benchmark developer-tools

Microsoft AI announced on September 4, 2026 that it is bringing the MAI-Image-2.6 image model to developers in Microsoft Foundry, and adding a faster MAI-Image-2.6-Flash alongside it¹.

Where Flash fits: aimed at latency-sensitive, high-throughput production workloads. Microsoft says it generates images 2.8x faster than GPT-Image-2-Medium while delivering 72% greater efficiency¹
Shared new capabilities: multi-reference editing, which brings together people, products, styles and scenes from different images; web grounding, which pulls in information from across the web; and dynamic aspect ratios that pick a format to suit the composition, with support for up to 1.5K resolution¹
Rankings: as of September 4, 2026, Microsoft states the model ranks No. 2 for both text-to-image and image editing on Arena, and on Artificial Analysis No. 2 for text-to-image and No. 1 for image editing¹

Both models can be tried in MAI Playground, or built with in Public Preview through Microsoft Foundry¹. No pricing appears on the page; Microsoft claims the combination of quality and efficiency “delivers the best price-per-Elo performance in the world”, without showing figures to support it¹. Microsoft has been adding purpose-specific MAI models, as with MAI-Code-1.1-Flash for coding. For image generation, developers now choose between maximum precision with MAI-Image-2.6 and production speed with MAI-Image-2.6-Flash¹.

Sources

Pushing the quality-cost frontier with MAI-Image-2.6 - Microsoft AI blog (September 4, 2026)