On September 9, 2026 (local time), OpenAI announced the image generation model "ChatGPT Images 2.5." This new model achieves more natural lighting and richer textures, while also improving adherence to editing instructions. Additionally, latency for image generation has been reduced by up to 50% compared to the previous "Images 2.0."

As a new feature, "Sketch" has been introduced, allowing users to use sketches drawn within the chat as references for images. By typing "@Sketch" in the chat box, rough shapes or scribbles can be converted into detailed images. Furthermore, features have been added that allow users to leave comments on specific parts of an image to provide editing instructions, as well as template functions to simplify the creation of posters and product photos.

"ChatGPT Images 2.5" is now being rolled out to all users of ChatGPT on desktop, mobile, and web, as well as ChatGPT Work and Codex. For developers, two new models are being provided via API: "GPT-Image-2.5 Flare," which prioritizes speed, and "GPT-Image-2.5 Sunburst," which offers higher precision.

Source: