OpenAI released GPT Image 2.6 on November 4, 2026, under the official name ChatGPT Images 2.6. The update brings animated GIFs, reference images you can reuse across chats, SVG export, image markup in the API and a new "complexity" control. This explainer covers what is new, how the two API models differ, how it compares with GPT Image 2.5, and how to turn the stills you make into real video clips.
What is GPT Image 2.6?
GPT Image 2.6 is the latest version of OpenAI's image generation model family. It powers image creation inside ChatGPT and Codex, and it is rolling out globally across every subscription tier on desktop, mobile and web. Developers get the same generation through two API models: GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst.
The release continues a fast cadence. GPT Image 1 arrived in March 2025, GPT Image 2 followed in April 2026, and GPT Image 2.5 shipped in September 2026 with Sketch, a feature that turns drawings into finished images. Version 2.6 keeps the Flare and Sunburst split introduced with 2.5 and focuses on motion, consistency and developer control.
OpenAI also says usage of image generation across its products has reached 4 billion images per week, which explains why each release now lands in both the consumer app and the API at the same time.
What's new in ChatGPT Images 2.6
Animated GIF creation
The headline feature is animated GIFs. You can create a short loop from a text prompt or from an uploaded photo. OpenAI positions it for characters, mascots, product shots, sketches, reaction images and storyboards. Once a GIF is generated you can control the animation speed, edit individual frames and share the result directly.
Reusable reference images
ChatGPT Images 2.6 lets you keep reference images of yourself, your pets or your recurring characters and reuse them later without uploading them again. For anyone building a series of images around the same subject, this removes the most tedious step in keeping a character consistent.
Image markup in the API
Markup, where you mark the precise area of an image you want changed, is now available in the API. Apps can let users circle or highlight a region and edit only that part instead of regenerating the whole frame.
SVG export
Images can be exported as SVG, which makes the output usable for websites, apps and print work where a scalable vector file is more practical than a raster image.
The complexity parameter
A new "complexity" parameter controls how much planning the model applies before it generates. OpenAI recommends low complexity for quick edits and medium or high complexity for scenes with many elements or constraints, such as a crowded poster or a layout with strict text placement.
GPT-Image-2.6 Flare vs Sunburst
The GPT Image 2.6 API offers two models tuned for different jobs. OpenAI claims both generate 50% faster than the previous versions.
| GPT-Image-2.6 Flare | GPT-Image-2.6 Sunburst | |
|---|---|---|
| Main strength | Speed | Detail and edit consistency |
| Best for | Professional layouts such as posters and product imagery | Detailed scenes and repeated edits of the same image |
| Quality sweet spot | Low quality for the fastest output | Medium and high quality |
| Speed note | As little as six seconds per image at low quality | Focused on quality and consistency rather than raw speed |
Start with Flare to iterate on layouts or variations fast, then switch to Sunburst to refine a chosen image with consistent edits.
GPT Image 2.6 vs 2.5: what changed
| Feature | GPT Image 2.5 (September 2026) | GPT Image 2.6 (November 2026) |
|---|---|---|
| Flare and Sunburst API models | Yes, introduced alongside Sketch | Yes, with OpenAI claiming 50% faster generation |
| Animated GIF creation | No | Yes, from prompt or photo, with frame editing and speed control |
| Reusable reference images | No | Yes, saved and reused without re-uploading |
| Image markup in the API | No | Yes |
| SVG export | No | Yes |
| Complexity parameter | No | Yes, low / medium / high |
If you already use GPT Image 2.5, the biggest practical gains are reference images for consistent characters and the complexity control for difficult compositions. GIFs are new territory entirely.
GPT Image 2.6 API: pricing and access
Both GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst are available in the OpenAI API. We have not verified per-image pricing for the new models, so see OpenAI's pricing page for current rates before you budget a project. Markup and the complexity parameter are part of the API release, so developers can expose the same targeted editing and planning controls that ChatGPT users get.
How to turn GPT Image 2.6 images into video
GPT Image 2.6 is excellent at stills and short loops, but a GIF is not a video. If you need a 5 to 10 second clip in 1080p with camera movement and audio, the best workflow is to generate a strong first frame in ChatGPT and animate it with a dedicated video model.
Step 1: Generate a still in ChatGPT
Write a prompt that describes a single, clear frame: one subject, one setting, one lighting setup. Leave room in the composition for movement, for example empty sky above a character who will jump, or space beside a product that will rotate. Use Sunburst-level quality (medium or high) when you plan to animate, since fine detail survives the motion better.
Step 2: Upload it to WandVideo image-to-video
Download the image and upload it to WandVideo image-to-video. The still becomes the first frame, and your text prompt describes what should happen next. Keep the motion prompt short and specific; our guide on how to write AI video prompts covers the shot-list structure that works best.
Step 3: Pick the right video model
Different models suit different shots. You can compare them all on the models page, but a quick starting point:
- Kling 4.0 for expressive character motion and smooth camera moves.
- Veo 3.1 for cinematic scenes and clips that benefit from generated audio.
- Seedance 2.5 for dynamic action and energetic, fast-moving shots.
Keep characters consistent across a sequence
Reference images are the key to character sequences. Save your character as a reference in ChatGPT Images 2.6, generate each scene's first frame with that same reference, and animate each frame separately on WandVideo. Because every clip starts from a frame with the same face, outfit and proportions, the cuts feel like one story rather than a set of strangers.
GIF or video model?
Use the built-in GIF feature when you need a quick loop: a reaction image, a mascot wave, a product spin for a chat message. Use a video model when you need a 5 to 10 second 1080p clip with audio for social media, ads or a landing page. Or prototype the motion as a GIF, then produce the final clip with a video model. Check pricing to see how many credits each model uses.
Prompt tips for GPT Image 2.6
Good image prompts for animation describe a frame, not a sequence. Name the subject, the setting, the light and the camera angle, and set complexity to match the scene.
Character still for animation (medium complexity):
A small orange fox mascot wearing a green scarf, standing on a mossy rock in a misty pine forest at dawn, soft backlight, three-quarter view, eye level, plenty of open space to the left, highly detailed fur.
Product poster (Flare, low complexity):
A minimalist poster of a matte black wireless earbud case on a pastel peach background, centered, soft studio shadow, bold headline space at the top, clean sans-serif layout, product photography style.
Cinematic first frame (high complexity):
A rainy neon-lit night market street seen from a low angle, a woman in a yellow raincoat holding a paper lantern in the foreground, steam rising from food stalls, reflections on wet pavement, shallow depth of field, 16:9 widescreen composition.
FAQ
When was GPT Image 2.6 released?
GPT Image 2.6 was released on November 4, 2026, under the name ChatGPT Images 2.6. It is rolling out globally in ChatGPT and Codex across every subscription tier, on desktop, mobile and web.
What is the difference between GPT-Image-2.6 Flare and Sunburst?
Flare is built for speed and professional layouts such as posters and product imagery, taking as little as six seconds per image at low quality. Sunburst delivers more detail and more consistent edits at medium and high quality, so it suits final assets and repeated refinements.
Is GPT Image 2.6 better than GPT Image 2.5?
It adds features 2.5 did not have: animated GIFs, reusable reference images, SVG export, image markup in the API and a complexity parameter. OpenAI also claims the Flare and Sunburst API models generate 50% faster than before.
How much does the GPT Image 2.6 API cost?
We have not verified the per-image pricing for GPT-Image-2.6 Flare or Sunburst. Check OpenAI's pricing page for current rates before planning a high-volume project.
Can GPT Image 2.6 make videos?
It can make short animated GIFs from a prompt or photo, with frame editing and speed control. For 5 to 10 second 1080p video with audio, generate a still in ChatGPT and animate it with a video model such as Kling 4.0, Veo 3.1 or Seedance 2.5 on WandVideo.
How do I keep the same character across multiple images?
Save the character as a reusable reference image in ChatGPT Images 2.6 and use it for every new scene instead of re-uploading. The same approach carries into video: animate each consistent still separately and the character stays recognizable across clips.
Sources: OpenAI: Introducing Images 2.6 · Wikipedia: GPT Image