OpenAI announced the release of ChatGPT Images 2.6 on November 4, 2026.
The company said Images 2.6 is now available in ChatGPT, Codex and the API, with ChatGPT and Codex availability across subscription tiers on desktop, mobile and web. OpenAI also said people create 4 billion images per week across ChatGPT, Codex and the API, putting the release into one of the company’s highest-volume consumer and developer use cases.
What changed in ChatGPT
For ChatGPT, Images 2.6 is designed to generate images faster and preserve details more reliably across edits, such as retaining a character’s clothing while changing the background. OpenAI’s Help Center says GIF creation is rolling out in ChatGPT. Users can ask for a looping animation, upload a still image or sketch and ask ChatGPT to animate it, then adjust playback speed and share the result.
The update also adds saved reference images. As the feature rolls out, a photo can be saved under a name and reused in later chats with an @ reference. For creators and small teams, this reduces repeated uploads for mascots, product concepts, pets or characters, but it also makes internal rules for consent and approved brand references more important.
OpenAI’s Images in ChatGPT help page says ChatGPT Images is available on all tiers across web, iOS and Android. It separately says Images with thinking is available on Plus, Pro and Business and is coming to Enterprise and Edu.
For developers, 2.6 splits speed and precision
In the API, the update arrives as two models: GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst. The company says Flare can generate images in as little as six seconds when quality is set to low, while Sunburst is intended for higher-quality image generation and stronger consistency across edits at medium or high quality.
| Model | Best fit | OpenAI’s positioning |
|---|---|---|
| GPT-Image-2.6 Flare | Fast drafts, quick comparisons and interactive creative tools | Optimized for response time and rapid iteration. |
| GPT-Image-2.6 Sunburst | Campaign creative, product images and detailed edits | Designed for more detail, layout accuracy and consistency across editing rounds. |
The API controls matter because speed and visual quality are now more separable. The new complexity parameter controls how much planning the model spends on a request, while quality controls visual detail and fidelity. OpenAI’s prompting guide recommends testing the matching 2.6 model against an existing workflow, keeping prompts, reference images, dimensions and quality consistent before changing one setting at a time.
For product or marketing teams, that creates a practical migration path: use Flare where latency is the constraint, use Sunburst where text placement, product shape, character consistency or final asset quality decides whether the image is usable.
GIF and SVG output expand the use cases
OpenAI’s API reference lists PNG, WebP, JPEG, GIF and SVG as output formats for image generation. For GIF output, the reference lists frame_count values from 2 to 16 and says GIF output does not support streaming, partial images, masks or custom output compression. SVG output is supported for generation, with the image model creating a reference image and a separate authoring model producing the SVG.
That distinction matters for production. GIF output is useful for lightweight loops, social reactions and animated mascots, but it should not be confused with full video generation. SVG output is better suited to scalable marks, interface concepts and graphics that may need editing after export, although generated SVGs still need review before use in brand or print work.
Pricing is token based
OpenAI’s developer pricing page lists the same U.S.-dollar GPT-Image-2.6 rates for Flare and Sunburst: $8 per million image input tokens, $30 per million image output tokens and $2 per million cached image input tokens. Text input is listed at $5 per million tokens, with cached text input at $1.25 per million tokens.
The company’s image-generation guide says markup-guided edits in the API use the same token rates as other edits with no extra fee. For SVG output, the total also depends on the authoring model’s token use. That makes cost testing important for teams using reference images, many variants or asset pipelines that retry until an image passes review.
Safety disclosures focus on GIFs, minors and moderation layers
OpenAI published a ChatGPT Images 2.6 system card on the same day as the release. The system card says the safety stack is based on the same foundations as ChatGPT Images 2.5, with added safeguards for GIF creation. It describes upstream policy checks before generation, downstream moderation of input images and generated outputs, and additional child-safety checks for ChatGPT image generation.
The system card also says GPT-Image-2.6 Sunburst and Flare were trained on datasets that include publicly available information, data from third-party partners and data provided or generated by users, human trainers and researchers. Those disclosures are relevant for organizations that have rules around copyrighted inputs, sensitive images or client-controlled assets.
OpenAI’s disclosures do not remove the need for review. The practical risk is that faster visual output can increase the number of assets needing approval, especially when images include faces, product labels, public figures, brand marks or motion. The safer workflow is still to treat generated assets as drafts until a person has checked accuracy, rights, policy fit and brand requirements.
I can suggest related Tech Help Canada articles based on the topic you’re reading now.
Tech Help Canada Staff researches, writes, and reviews practical content for business owners and professionals. Our coverage spans business, marketing, SEO, technology, and the tools and systems people use to grow and operate online. We focus on clear, useful information backed by research, hands-on experience, and editorial review. Learn more about our team and editorial standards. Need help with something? Contact Us
