Microsoft's MAI-Image-2.6 Rises to No. 2 on Arena Leaderboard with 1,336 Score
Microsoft has released MAI-Image-2.6, its latest text-to-image generation model, which climbed to second place on the Arena leaderboard with a score of 1,336, up about 79 Elo points from the previous MAI-Image-2.5. The model surpassed offerings from Meta, Google, ByteDance and xAI, trailing only OpenAI's GPT-Image-2 by 45 points. Microsoft said the model delivers breakthroughs in text rendering, portrait drawing and 3D image generation, and supports multi-reference image collaboration. The model is now available on Arena, with integration into MAI Playground and Microsoft Foundry planned this week.
Microsoft has officially released its new text-to-image generation model, MAI-Image-2.6. According to the latest Arena leaderboard data, the model quickly rose to second place in the text-to-image category after launch, surpassing multiple mainstream models from competitors including Meta, Google, ByteDance and xAI, and now trails only OpenAI's GPT-Image-2. Microsoft launched its first self-developed text-to-image model, MAI-Image-1, in October 2025, debuting at ninth place; in March 2026, it released the second-generation upgrade MAI-Image-2, which entered the top three on the strength of improved image quality and performance; in May of the same year, MAI-Image-2.5 continued to hold a top-three position; and last month, Microsoft also introduced the MAI-Image-2.5-Pro model for high-precision image generation, image editing and complex text rendering requirements.
The newly released MAI-Image-2.6 scored 1,336 points in the Arena text-to-image category, an improvement of about 79 Elo points over the previous MAI-Image-2.5. Mustafa Suleyman, head of Microsoft AI, said on social media that sustained technical efforts enabled the model to surpass comparable competing products including Nano Banana, Meta and Grok. On the leaderboard, MAI-Image-2.6 trails the top-ranked OpenAI GPT-Image-2 (medium version) by 45 points, while leading third-placed Grok Imagine Image 2.0 (low version) by 20 points.
In terms of technical capabilities, Microsoft said MAI-Image-2.6 has achieved breakthroughs in text rendering, portrait drawing and 3D image generation, producing images with greater commercial polish and realism, and performing strongly in application scenarios such as product display, brand promotion and cinematic-style creation. The new model supports collaborative creation based on multiple reference images, offers stronger visual grounding, and provides greater autonomy in inference control, output format and resolution adjustment.
Users can currently experience the MAI-Image-2.6 model on the Arena platform. Microsoft plans to disclose more technical details of the model in the coming weeks, and will progressively integrate the model into MAI Playground, Microsoft Foundry and its other products and services this week.
Why this event matters
The event has a measured impact on 1 industry. The strongest current signal is mixed for Artificial Intelligence, with intensity 60/100 and 80% confidence over a short term horizon.
Impact figures are analytical estimates that combine direction, intensity, confidence and event importance. They are not investment advice.