Affiliate links on Android Authority may earn us a commission. Learn more.
Sep 9, 2026 — 2:48 AM ET
- OpenAI has released the new ChatGPT Images 2.5 model, which can understand complex instructions and has better visual intelligence.
- It can also maintain distinct details from reference images and precisely edit a single element in photos.
- ChatGPT Images 2.5 is also better at editing over multiple steps without degrading quality.
People have been creating and editing images with ChatGPT for a long time. So much so, in fact, that they’ve occasionally stumbled upon absolutely bizarre generations. However, OpenAI is now making it easier and faster to generate images in ChatGPT, while also making them more realistic and giving users some advanced tools for accurate editing.
OpenAI announced the release of ChatGPT Images 2.5 as its new state-of-the-art image generation model. According to OpenAI’s blog post, the new model creates images with more natural lighting and textures without obliterating the subjects in an uploaded image. The company also claims that the new model is 50% faster at creating images.
ChatGPT Images 2.5 also adds several new features that give users more control over the entire process. A new “Sketch” feature lets users draw a reference image inside ChatGPT, while built-in templates can make it easier to design flyers and posters. I can only imagine the sheer amount of AI slop posters that are about to flood Instagram in the coming weeks.
The new model also better preserves details such as faces from reference photos. It can carry over distinct features from reference images across new settings and visual styles. Additionally, ChatGPT Images 2.5 can precisely edit a single element in photos. This means users can now change the background without altering the subject or edit the text in an image while keeping the rest of the image intact. It can also create images with transparent backgrounds, which is great news for anyone looking to create stickers with ChatGPT.
OpenAI has also improved multi-turn editing consistency in the new model. It is now more likely to follow editing instructions reliably across multiple edits without degrading image quality or altering earlier changes. ChatGPT Images 2.5 also features improved visual intelligence and can understand complex instructions and real-world information to generate better images. It also lets users comment directly on images for more precise edits and share their prompts with others, along with the generated images.
The new image model is rolling out to all users on ChatGPT, ChatGPT Work, and Codex on the web, desktop, and mobile.
Thank you for being part of our community. Read our Comment Policy before posting.