OpenAI launched the latest model ChatGPT Images 2.0 this morning which can be seen as their answer to Gemini Nano Banana. Images 2.0 can generate images that look more realistic, understand context better, support resolutions up to 2K and support image generation in various screen ratios.
A more interesting and useful feature is that Images 2.0 can perform an internet search first to generate more accurate images based on the given prompt. For example, when the prompt to generate a poster of an item sold at store A is given, Images 2.0 will first search for what the item is before generating the requested poster. Users do not need to upload images to be used as a source as requested by Gemini.
OpenAI also said that Images 2.0's training data is also up to date until December 2025. So the generation will be more up-to-date without errors. In addition, Images 2.0 can also generate images with non-English alphabets better.
The majority of generative AI fails to generate readable words when required to use the Jawi, Chinese, Japanese, Korean, Hindi and Russian alphabets for example. With an understanding of foreign languages and alphabets, these errors can be avoided.
Images 2.0 is also trained to produce images with more diverse visual styles, generate comics of more than one page, and produce complete images with safe sizes for professional printing.
