London — May 30, 2026
Recraft, an independent AI lab, today announced that Recraft V4.1 Utility Pro has entered the Artificial Analysis Image Arena as the third-ranked lab worldwide and the highest-ranked text-to-image model from any lab not part of a Big Tech parent.
The Artificial Analysis Image Arena ranks text-to-image models by Elo, based on blind human preference votes. On the current leaderboard, the top five entries by Elo are GPT Image 2, GPT Image 1.5, Nano Banana 2, Nano Banana Pro, and Recraft V4.1 Utility Pro. The four models above Recraft belong to two labs: OpenAI and Google. Recraft is the third lab on the board.
It is also the first independent. V4.1 Utility Pro outranks every other text-to-image model on the benchmark, including those from xAI, and ByteDance and Microsoft.
Recraft's research team is fewer than 20 people.
"We set out to build models optimised for aesthetics, to go beyond prompt accuracy," said Anna Veronika Dorogush, Founder and CEO of Recraft. "This result shows that high visual value is important for everyone, not just professionals. The fact that we achieved such a high rating with x100 less resources than everyone else on the list makes me incredibly proud of the team.”
The V4.1 family covers three modes:
Learn more about the models in the release blog: https://www.recraft.ai/blog/recraft-v4-1-more-beautiful-by-nature
Recraft builds AI models optimised for aesthetics. Recraft's models power Recraft Studio, the application used by creative teams worldwide to produce visuals at a professional standard, and are available through the Recraft API for developers and enterprise customers.
.svg)