Microsoft's New MAI-Image-2.7 Model Climbs to the Top of AI Image Editing Charts

Microsoft has introduced MAI-Image-2.7 Preview, its latest in-house AI image model, which debuted in second place on Artificial Analysis’s image editing leaderboard, ahead of OpenAI’s GPT Image 2.5 Flare.

The new model also secured third place on Arena’s image editing leaderboard, with Microsoft highlighting improvements in complex editing, image restoration, lighting adjustments, and text manipulation.

MAI-Image-2.7 Preview outperforms GPT Image 2.5 Flare

According to Microsoft’s announcement, MAI-Image-2.7 Preview achieved an Elo score of 1,193 on Artificial Analysis, placing it just behind OpenAI’s GPT Image 2.5 Sunburst, which scored 1,199.

Meanwhile, GPT Image 2.5 Flare scored 1,182, while Microsoft’s previous MAI-Image-2.6 model reached 1,151.

This represents a 42-point improvement over MAI-Image-2.6 and puts Microsoft’s latest model within six points of the leaderboard’s top performer.

Microsoft previously introduced MAI-Image-2.6 in August 2026, followed by the faster and more affordable MAI-Image-2.6-Flash in September.

Microsoft improves complex editing and image restoration

Microsoft claims MAI-Image-2.7 Preview outperforms its predecessor across all 10 use cases and seven editing actions evaluated by Artificial Analysis.

The biggest improvements include image enhancement and restoration, allowing the model to improve image quality and recover visual details.

It also handles reasoning-based edits that require understanding objects and their surroundings, while offering better control over lighting, environments, and visual styles.

Microsoft says the model can also modify text and symbols within images while keeping them naturally integrated into the scene.

These improvements focus on giving users greater control over existing images, particularly when editing complex scenes or making precise visual changes.

MAI-Image-2.7 Preview is available for testing

Microsoft has made MAI-Image-2.7 Preview available for image editing through MAI Playground, Arena, and Artificial Analysis.

Unlike MAI-Image-2.6, the new preview initially focuses on image editing rather than generating images from text prompts. Microsoft says text-to-image generation support will arrive soon.

The company has not announced pricing or availability through Microsoft Foundry.

Microsoft previously brought MAI-Image-2.6 and MAI-Image-2.6-Flash to general availability in Microsoft Foundry, but it remains unclear when developers will gain access to the newer MAI-Image-2.7 model through the platform.