MAI Image 2.7
MAI Image 2.7 is Microsoft AI’s image generation and editing model, announced at Microsoft Ignite on 2026-11-17 and listed in Microsoft Foundry. It is a diffusion model that generates images from text and edits existing images, aimed at creative and production work. Part of the MAI series.
What it does (Foundry catalogue card)
- Input: text and images. Output: images.
- Text-to-image: photorealistic and cinematic images, high-fidelity portraits, and product, branding and marketing visuals.
- Editing: object removal and replacement, attribute changes, inpainting, text updates, and artefact clean-up such as motion blur, while keeping the composition.
- Text in images: improved rendering of labels, posters, packaging and signage.
- Consistency: visual consistency across repeated edits.
- Training: a flow-matching objective.
Microsoft’s claims
- Ignite headline: “#1 on Artificial Analysis image editing.”
- The Ignite post’s headline positions it as “frontier quality at a fraction of the cost”. The cost figure is not given in the text reviewed.
Limits and specifications
- Context window 32,000 tokens; output limit 4,096 tokens.
- Lifecycle: Preview. Behaviour and availability may change.
- Hosted as a “Direct from Azure” model; version 2026-11-02.
- The catalogue card gives no limits beyond the token figures above.
Lineage
MAI-Image-2 launched in April 2026 with the first generation. Microsoft’s September 2026 post on MAI Image 2.6 is the step before this version. See MAI series for the full timeline.
Sources
- Microsoft Foundry catalogue: MAI-Image-2.7 (fetched 2026-11-22)
- Microsoft AI: MAI Image 2.7 (2026-11-17), headline and claims only; body not read in full