OpenAI has rolled out Images 2.6, the latest version of the image generator in ChatGPT, Codex and its API. In ChatGPT and Codex, the new model is already available worldwide on every plan — on the web and in the desktop and mobile apps.
Images 2.6 generates images 50% faster than its predecessor. Developers get two versions of it in the API. The first, GPT-Image-2.6 Flare, is built for speed: at the low quality setting it can produce an image in as little as six seconds. Flare also does a better job of laying out posters, product shots and other professional visuals.
The second, GPT-Image-2.6 Sunburst, targets work that demands detail and precision. At medium or high quality, it delivers better results and keeps characters, products and brand elements more consistent across successive edits. More precise editing was already the focus of the previous version, Images 2.5, which OpenAI released in September.
The headline feature of Images 2.6 is animated GIFs. OpenAI is pitching them mainly at marketers and content creators. They can bring a brand mascot to life for a campaign, make a meme for social media, or build an animated storyboard to work out a character’s movements before video production begins. To do this, the model generates a set of sequential frames — a sprite sheet — that loops into an animation. Even a rough hand-drawn sketch can serve as the starting point.
Follow Mediasat on Google News: the latest in TV, telecom and tech
Ahead of launch, photo editor Picsart and AI content platforms fal and Higgsfield tested Images 2.6 in early access. AI art platform OpenArt used it to create sample GIFs. The model is also coming to Figma Weave.
For developers, the update also expands the editing toolkit. Image markup, previously available only in ChatGPT, is now in the API, letting teams pinpoint exactly where a change is needed. Finished images can also be exported as SVG, a vector format ready for websites, apps and print.
Another API addition is a complexity parameter. It controls how much planning the model does on a scene before generating it. Low settings suit quick edits, while medium and high work better for images with many elements or constraints. Unlike the quality setting, which governs detail, complexity determines how much goes into the scene. For most edits, OpenAI recommends pairing low complexity with whatever quality level the project requires.
According to the company, users create 4 billion images a week across ChatGPT, Codex and the API — more than half a billion a day on average. OpenAI has published API pricing for Flare and Sunburst separately.
