Arena.ai rolls out detailed categories for image editing models, and GPT Image 2 is running away with it

Arena.ai just changed that. The community-voted benchmarking platform has introduced detailed prompt categories for evaluating image editing models, breaking performance down across seven core dimensions. And the early results are pretty decisive: OpenAI’s GPT Image 2 leads in every single one.

The numbers behind the dominance

GPT Image 2 currently sits atop the Arena.ai Image Edit leaderboard with an Elo score of 1463, plus or minus 4 points. That ranking is based on more than 165,000 community votes.

The model, which launched in April 2026, didn’t just edge out the competition. In some categories, it holds a lead of 242 Elo points. For context, in chess Elo ratings, a 200-point gap typically means the higher-rated player wins about 75% of the time.

The seven new categories join additional evaluation metrics that Arena.ai tracks, including prompt adherence, aesthetics, and photorealism.

Who else is in the race

Behind GPT Image 2, Meta’s Muse Image holds the second-place position, with Microsoft’s MAI-Image-2.5 rounding out the top three. Google’s Gemini variants and ByteDance’s Seedream models also appear in the rankings.

The broader platform has accumulated 28.5 million total votes across 52 models in the image editing arena.

Arena.ai first introduced detailed prompt categories for text-to-image evaluations back in February 2026, using clustering analysis of user prompts to identify the most meaningful ways to slice performance data. The expansion to image editing follows the same methodology.

What this means for the AI market

OpenAI’s sweep across all seven categories reinforces its position at the top of the generative AI hierarchy. The company has now achieved first place in both text-to-image and image editing tasks on Arena.ai.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.