OpenAI launches ChatGPT Images 2.0 with improved text rendering to rival Nano Banana 2

OpenAI has launched ChatGPT Images 2.0, a new image generation model positioned to take on Google’s Gemini Nano Banana 2. The updated model puts particular focus on one of the most persistent weaknesses in AI image generation: text. According to OpenAI, Images 2.0 is much better at rendering readable, accurate text and preserving character clarity across more complex visuals, including icons, chat interfaces, and other UI heavy compositions, which means your generated images should finally stop being filled with those weird fake characters in text.

OpenAI is releasing the model in two versions: ChatGPT Images 2.0 instant and ChatGPT Images 2.0 thinking. The instant version is the standard model available to all ChatGPT and Codex users, while the thinking version is reserved for paid subscribers and adds built in reasoning, real time web lookups, multiple outputs from a single prompt, and self checking for more advanced image generation.

The update also adds support for up to 2K resolution and a wider range of aspect ratios, including both very wide and very tall formats. OpenAI also highlights stronger multilingual rendering, especially for Japanese, Korean, Chinese, Hindi, and Bengali. For developers, the underlying gpt-image-2 model is now available through the API.