ChatGPT Images 2.6: GIFs, Saved References and Faster Images
OpenAI's Images 2.6 makes animated GIFs, remembers your reference photos and works faster. Here is what changed in ChatGPT and in the API.
Illustration made on Lumeta with GPT Image 2.5Make one like it
OpenAI released Images 2.6 on November 4, 2026, two months after Images 2.5. It brings three things creators will notice right away: ChatGPT can now make animated GIFs, it can keep your reference photos so you do not upload them every time, and OpenAI says images come out 50% faster than before. The update is in ChatGPT, in Codex and in the API, where it arrives as two models, Flare and Sunburst.
What is new in ChatGPT
- Animated GIFs. Turn a prompt or an uploaded picture into a short animation, change its speed, and edit single frames.
- Saved references. Save a photo of yourself, your pet or a character once, then reuse it in new images. OpenAI has not published how many you can keep on each plan.
- Speed. OpenAI says Images 2.6 is 50% faster. It has not said how it measured that.
- Every plan. The update reaches all ChatGPT plans, Free included, with higher limits on paid plans. Image creation with Thinking stays a Plus and Pro feature.

Illustration made on Lumeta with GPT Image 2.5Make one like it
Flare and Sunburst: the two API models
Developers get two models. OpenAI describes gpt-image-2.6-flare as the one for fast iteration and gpt-image-2.6-sunburst as the one for detailed visual work. Both can return an animated GIF or an SVG vector file, both accept markup to guide an edit, and both add a new complexity setting.
- Complexity controls how much the model plans before it draws:
low,mediumorhigh, withhighas the default. Low suits quick, focused edits. High helps with dense text, busy layouts and scenes where objects relate to each other. It costs nothing extra, though higher settings can be slower. - GIFs have 2 to 16 frames (9 by default), each shown for 20 to 2,000 milliseconds (200 by default).
- SVGs come one at a time from the generation endpoint only, and cannot be used as input for edits.
How much it costs
OpenAI kept the token prices of Images 2.5. On OpenAI's pricing page, checked November 16, 2026, both 2.6 models cost $5 per million text input tokens, $8 per million image input tokens and $30 per million image output tokens. OpenAI notes that the models can use different numbers of tokens for the same quality setting, so the price of one picture can differ from 2.5 even though the rates match. Its image guide has a calculator for estimates.
Try GPT Image 2.6 on Lumeta
You can make and edit pictures with OpenAI's model on Lumeta's GPT Image 2.6 page. Add up to 10 reference images, paint over the area you want changed, and pick the shape and size, up to 4K.
GPT Image 2.6Make posters, product shots and photo edits with OpenAI's image modelWhen did ChatGPT Images 2.6 come out?
OpenAI released it on November 4, 2026, in ChatGPT, Codex and the API.
Can ChatGPT make GIFs now?
Yes. With Images 2.6 you can turn a prompt or a picture into an animated GIF, change its speed and edit single frames.
Is Images 2.6 free?
It is on every ChatGPT plan, including Free, but free users get fewer images than paid plans.
What is the difference between Flare and Sunburst?
In the API, Flare is built for fast iteration and Sunburst for detailed work. They share the same token prices and the same features, including GIF and SVG output.


