The benchmark that mattered most — Hugging Face's Text-to-Image Leaderboard by Artificial Analysis — had been comfortably occupied by the usual suspects: Midjourney, DALL·E 3, Ideogram. Then in October 2024, a model from a 50-person London startup displaced all of them with an ELO rating of 1172, and held the #1 position for five consecutive months. The model was Recraft V3. The startup was Recraft. And behind it was Anna Veronika Dorogush — a former professional model turned mathematician turned machine learning engineer turned founder — who had spent the previous two years quietly building the most technically sophisticated AI image model the design industry had seen.
What Dorogush built is not an image generator with brand features bolted on. It is a design platform that was conceived from the first line of code around the specific, painful problem that every designer and marketing team eventually hits: you can generate a hundred beautiful images, but none of them look like each other, and none of them look like your brand. Recraft V3 solves this with positioning control — the ability to specify exactly where a logo, product, or text element should appear in a generated image, not approximately, not suggested by the prompt, but precisely. Combined with custom brand style training (upload your visual assets once, generate on-brand imagery forever), native SVG vector export (true scalable vectors, not raster images converted to SVG), and a real-time collaborative infinite canvas, the result is a platform that Amazon, NVIDIA, Salesforce, Uber, Ogilvy, Netflix, HubSpot, Airbus, and Asana now rely on for production creative work. $42M raised from Khosla Ventures, Accel, Madrona, Nat Friedman, Elad Gil, Dharmesh Shah (HubSpot CTO), Akshay Kothari (Notion co-founder), and John Maeda (Microsoft CVP of Design). 4 million users. 700% user adoption growth in one year. And a founder who, when asked what lesson she carries from her years as a professional model before entering tech, said: "Grinding isn't everything. You have to be excellent at what's mission-critical."
