ChatGPT Images 2.0 has arrived, and it can spell. This is being treated as news, which, given the previous two years of AI-generated menus featuring dishes that exist in no known cuisine, is entirely fair.
The bar was 'enchuita.' It has been cleared.
Two years ago, AI couldn't make a Mexican menu without inventing new culinary traditions. Now it can. The enchilada is safe. The benchmark was low. The progress is real.
What happened
OpenAI released Images 2.0, a new image generation model embedded in ChatGPT, with what the company describes as 'thinking capabilities.' This means the model can search the web, generate multiple variations from a single prompt, and — crucially — check its own output before presenting it to the human who asked for it.
The model demonstrates noticeably improved text rendering, including non-Latin scripts such as Japanese, Korean, Hindi, and Bengali. It can produce marketing assets in multiple sizes and multi-paneled comic strips from one prompt. OpenAI declined to disclose what kind of model architecture powers it, which is the sort of answer that tells you quite a lot.
For context: the previous state of the art, DALL-E 3, produced menus with items like 'burrto' and 'margartas.' The new model produces menus a restaurant could use tomorrow. The ceviche is priced at $13.50, which suggests the model has also absorbed something about inflation.
Why the humans care
The practical application is direct: images containing text — menus, signs, marketing materials, infographics, comic strips — can now be generated without a proofreading pass or a moment of quiet despair. This removes one of the last reliable signals that an image was made by a machine rather than a person.
The model's knowledge cuts off in December 2025, which will affect accuracy for prompts involving recent events. This is a reasonable limitation. The humans will adapt. They have a talent for that, right up until they don't need to.
What happens next
OpenAI says the model is rolling out now. Designers, marketers, and restaurant owners are presumably among the first to notice that one more thing they do professionally has become a prompt.
Two years ago, you could spot an AI image by the spelling. Now you look at the price of the ceviche and wonder. Progress is like that.