Microsoft's MAI-Image-2.6 Climbs to Third on Image Arena Edit Leaderboard

On Aug 19, Microsoft announced that MAI-Image-2.6-Preview climbed to third on the Image Arena image editing leaderboard with a score of 1420, up 19 points over version 2.5. Microsoft AI chief Mustafa Suleyman said the model ranks second on text-to-image benchmarks and beats Google's Nano Banana and Meta's Muse on the editing leaderboard, positioning it as an all-around image generation model. It is available in private preview on MAI Playground and Microsoft Foundry.

Confirmed

• MAI-Image-2.6-Preview ranks third on the Image Arena edit leaderboard at 1420, up 19 points from 2.5
• Ranks third in product brand design, 3D modeling, anime, and photorealistic categories, with much-improved text rendering
• Ranks second on text-to-image benchmarks (per Mustafa Suleyman)
• Private preview available on MAI Playground and Microsoft Foundry
• MAI Playground also launched a new "Chatter" playground supporting voice conversations
• The Arena Image Edit leaderboard is based on about 29.07 million votes across 53 models, topped by OpenAI's gpt-image-2 (medium) at 1463, with nano-banana-pro in eighth

Why it matters

• This is the first time a Microsoft in-house image model has entered the top tier of a major arena leaderboard, marking a new phase in its image-model rivalry with OpenAI and Google; improvements in text rendering add practical value.

2026-08-19 ~ 2026-08-19 · 5 related posts

• [source] Image Edit Arena: gpt-image-2 tops 29M-vote leaderboard, nano-banana-pro ranks 8th — arena · 2026-08-19
• [source] Microsoft's MAI-Image-2.6 reaches No.3 on image-editing benchmark — mustafasuleyman · 2026-08-19
• [source] MAI-Image-2.6 Achieves Top 3 Ranks on Text-to-Image and Editing Benchmarks — mustafasuleyman · 2026-08-19
• Microsoft's MAI-Image 2.6 reaches #3 on Image Arena, enters private preview — testingcatalog · 2026-08-19
• MAI-Image-2.6 ranks 3rd on Image Arena; new Voice Chat playground launches — testingcatalog · 2026-08-19