Microsoft launches MAI-Image-2.6: matches GPT-Image-2 on edit Elo, Flash runs 2.8x faster at a fraction of the price
xiaohu · x · 2026-09-05
Microsoft released MAI-Image-2.6 and an optimized MAI-Image-2.6-Flash, now in public preview via Microsoft Foundry, with a MAI Playground for casual users.
Key capabilities
- Multi-image reference editing: combine product, person, material and scene references into one image
- Web-grounded generation using web pages and documents as context
- Local edits and aspect-ratio adaptation, output up to 1.5K resolution
Key numbers
- Image editing Elo: 2.6 scores 1324, just 6 points below GPT Image 2 (high) at 1318; Flash scores 1311 (evaluation uncertainty 9-10 points)
- Pricing: $38.9 per 1,000 images for 2.6, $19.5 for Flash, vs $211 for GPT Image 2 (high)
- Speed: in Microsoft internal tests (1024×1024, 100 RPM), Flash's median response is 12.2s vs 33.6s for GPT-Image-2-Medium — roughly 2.8x faster
- Rankings: 2.6 places second on both Arena text-to-image and image-edit leaderboards, and second/first on Artificial Analysis
The flagship 2.6 targets quality and fine editing, while Flash targets speed and cost for batch generation.
More from Multimodal
- This stunning W51A star-forming region image was AI-processed from raw JWST captures — fofrAI · 2026-10-21
- Purify Before You Align: filtering locally predictable info improves diffusion alignment — volokuleshov · 2026-10-21
- Mistral AI launches Voxtral 3, a real-time voice agent stack in 50+ languages — The Decoder · 2026-10-21
- Full AI Virtual Influencer Character Sheet for 'JataJett': Turnarounds, Expressions and Action Modes in One Page — CurieuxExplorer · 2026-10-21
- AI satire video: 'Paris City Tours, 50% off — slightly different experience' — DavidmComfort · 2026-10-21
- AI-Generated Series 'The Sage' Hits Episode 8, Made Entirely with Invideo Agent — Kavanthekid · 2026-10-21