2026.11.10

GPT Image 2.6 is out: a roundup of the model's features

Image generation is 50% faster than before. It also supports GIF animation generation, editing by specifying particular parts of an image, and SVG output. In the API, you can choose between Flare, which prioritizes speed, and Sunburst, which prioritizes quality.

GPT Image 2.6 is out~!! I put together a roundup of the model’s features, so I’m sharing it!

Image generation AI is getting to the point where it doesn’t just create images but can handle everything from “fast generation → partial edits → animation → use in actual production work.”

With Images 2.6, image generation is 50% faster than before. It also supports GIF animation generation, editing by specifying particular parts of an image, and SVG output. On top of that, in the API you can choose between Flare, which prioritizes speed, and Sunburst, which prioritizes quality, so you can use it in a way that fits your production workflow.

More in the replies!

https://x.com/suna_gaku/status/2118137297711616260

① Image generation AI is evolving from “make it once” toward “edit it repeatedly” Images 2.6 emphasizes not only generation speed but also keeping the characteristics of characters, products, and brands intact even through repeated edits. It’s suited to a production flow where you improve existing designs instead of recreating images from scratch every time.

② Image generation is 50% faster than before. You can choose models by speed and quality The API offers GPT-Image-2.6 Flare and Sunburst. Flare prioritizes speed and can generate in as little as 6 seconds with the low quality setting. Sunburst, at medium / high quality, emphasizes high-quality image generation and consistency between edits. You can choose the model depending on the use case.

③ GIF generation lets you create everything from still images to animations It supports brand mascots, social media reactions, looping animations of product images, and more. You can express a character’s movements across multiple frames, which can also be used for checking motion before video production and for storyboards.

④ You can create animations even from simple hand-drawn sketches Even without a finished design, you can generate a moving character based on a hand-drawn illustration. The article introduces an example of animating a sketch of an owl. It makes it easier to develop early ideas into expressions with movement.

⑤ Image Markup lets you specify exactly where to edit an image Image Markup, which was available in ChatGPT, now also works in the API. Because you can specify which part of the image to change, you can clearly convey the part you want to fix instead of recreating the whole thing. It’s also easy to use when building image editing into apps and production tools.

⑥ SVG output lets you connect generated images to actual production work Now that it can output in SVG format, you can use the results in websites, apps, printed materials, and more. It’s not just about generating visuals with AI; it also becomes easier to carry them through to the design work and implementation that follow.

⑦ Adjust Quality and Complexity separately Quality adjusts the detail and fidelity of the image, while Complexity adjusts the amount of planning that goes into composing the scene. For simple edits, use low complexity to speed things up, and for scenes with many elements and constraints, choose medium / high complexity. The key is to think about quality and processing complexity separately.

⑧ You can build image generation into a continuous production flow rather than treating it as a one-off task The article also introduces uses in Figma Weave, Picsart, and others. What matters is being able to repeat generation, editing, and reuse while maintaining brand consistency. Images 2.6 is available in ChatGPT, Codex, and the API, so you can use it to fit your production environment.

https://x.com/suna_gaku/status/2118137300223926462