OpenAI released gpt-image-2 on April 21, and within 12 hours it had claimed the #1 spot on the Image Arena leaderboard by a +242 point margin — the largest lead ever recorded on that leaderboard. The headline feature is text rendering: the model now generates print-ready menus with accurate pricing, multilingual labels (Japanese, Korean, Chinese, Hindi, Bengali) without warped letters, and embedded working QR codes. It's also the first image model with native reasoning ("Thinking" mode) that lets it plan layouts before rendering, and outputs are 4K-resolution.
What makes this release different is the marketing-grade photorealism with deliberate imperfections — exactly the kind of output you'd send a creative team for review and not flag as AI. Real demos circulating this week include: a six-slide marketing deck for a fictional energy drink generated as a single prompt, restaurant menus that hold up under dense layouts, e-commerce product shots, branded campaign assets, and editorial-quality lifestyle imagery. A UCO Bank check generated by Images 2.0 went viral — and triggered fraud concerns severe enough to prompt fact-checks across Indian financial outlets. The Twitter chorus is back with the line "Graphic designers are cooked," which is an exaggeration, but the shift this time feels meaningful in a way prior generations didn't. For marketing teams making collateral at scale — campaign visuals, social assets, product photos — the question is no longer "can AI replace this?" but "how much real photography do we genuinely still need?"
Live demo: ChatGPT Images 2.0 generating marketing collateral