OpenAI released GPT Image 2 on April 21, a rebuilt image generation model that replaces GPT Image 1.5 in ChatGPT and in the company’s developer API. The company says the new model was redesigned from the ground up rather than incrementally tuned, with the clearest improvements in rendering legible text and producing higher resolution images.
Text that reads correctly
Garbled lettering has long been one of the most visible weaknesses of AI image generators. Signs, labels and headlines in generated pictures often came out as near-words or scrambled characters, limiting the technology’s usefulness for design work. OpenAI claims GPT Image 2 renders requested text accurately more than 99 percent of the time, and reports from developers who gained early access through API platforms such as fal.ai describe strong results on posters, infographics and multilingual text.
That figure is OpenAI’s own measurement, and independent evaluations are still being compiled. Early comparisons on public image arenas, where users vote between outputs from different models without knowing which is which, placed GPT Image 2 at or near the top of the rankings in the days after launch, according to TechWyse.
Resolution and formats
The model can produce images at resolutions approaching 4K, a substantial step up from its predecessor, and supports a wider range of aspect ratios, including 16:9 widescreen output suited to presentations, video thumbnails and advertising. OpenAI positions these changes as making the model viable for professional production workflows rather than only for casual experimentation.
The model is available in ChatGPT for paid users, through the OpenAI API, and in Codex, the company’s coding agent, where it can generate assets as part of software projects.
A crowded field
The release lands in an increasingly competitive market. Google has pushed its Gemini-based image models, popularly known by the “Nano Banana” nickname, toward photorealism and precise editing, and Midjourney continues to attract designers with its distinctive aesthetic output. Each new release from one company has tended to prompt a response from the others within weeks.
For OpenAI, GPT Image 2 is also a strategic product. Image generation was one of ChatGPT’s largest growth drivers in 2025, when a previous model’s launch produced a surge in sign-ups, and the company has an interest in keeping its image tools competitive as consumers increasingly judge AI assistants by their creative capabilities as well as their text.
