OpenAI Releases Images 2.0, a More Advanced ChatGPT Capable of Understanding Context and Accurately Generating Images

OpenAI has officially released a new image generation feature in ChatGPT, named ChatGPT Images 2.0, on Tuesday (21/4/2026). This update represents a significant leap from its predecessor, ChatGPT Images 1.5 (released in December 2025). In its official statement, OpenAI states that this model delivers unprecedented levels of detail and accuracy. The model is evaluated as superior in following instructions, maintaining details, and rendering small elements such as text, icons, and complex visual compositions, with resolutions up to 2K. ChatGPT Images 2.0 no longer simply “draws” based on commands. The model can now analyse, plan, and compose visual structures before generating images, resulting in more complex and contextual outcomes. These new capabilities allow users to create various types of visual content that were previously difficult for AI image generators to achieve. From complete infographics and presentation slides to maps and multi-panel manga comics, all can now be produced with a single command. Nevertheless, OpenAI notes that the model’s knowledge base remains limited to December 2025. This means that for requests related to recent events, the generated results may not be entirely accurate. Another notable improvement is in the ability to display text within images. Previously, text in AI-generated images often appeared random or illegible; now, ChatGPT Images 2.0 can produce neat and clear typography, even in complex compositions such as posters or scientific diagrams. Moreover, the model supports various languages more accurately, including non-Latin characters such as Japanese, Korean, Chinese, Hindi, and Bengali.