OpenAI has announced the release of a new model for generating and editing visual content, ChatGPT Images 2.5. The update speeds up the image creation process, better reproduces details, and adds new tools for controlling graphics.
OpenAI has announced the release of a new model for generating and editing visual content, ChatGPT Images 2.5. The update speeds up the image creation process, better reproduces details, and adds new tools for controlling graphics.
The company’s press service reported on the release. Along with the new version, developers gained access to two specialized API models — GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst. The updated system is already being deployed for users of ChatGPT, ChatGPT Work plans and developers in Codex in the web version, on desktops and mobile devices.
The key technical change was a 50% increase in generation speed compared to version 2.0. The algorithms have improved visual intelligence, better understanding complex text instructions and real-world information, and reproducing natural lighting and textures.
New features include the ability to create images with transparent backgrounds, which simplifies the development of stickers and visual elements. The system also allows you to make point changes to individual fragments — for example, redraw the background or edit text without changing the main object. The model is better at recognizing key features from reference photos and transferring them to new compositions or styles without distortion, and is also better at maintaining context during long-term editing in chat.
The user interface now includes a Sketch tool for creating initial sketches directly in the dialog box using the @Sketch command. Also added are templates for posters and promotional materials, the ability to leave comments on specific areas of the image for precise editing, and a tool for sharing prompts.
OpenAI has offered two versions of the API for third-party services. The Flare model is optimized for high-performance systems, rapid prototyping, and social media content generation. The Sunburst version, on the other hand, is focused on complex commercial design, where maximum accuracy of detail is a priority.
OpenAI continues to use built-in C2PA metadata and invisible watermarking to protect content. According to the company, over 3 billion images are created each week through the ChatGPT ecosystem.
