On November 4, 2026, OpenAI introduced Images 2.6, an updated image-generation system for ChatGPT, Codex, and the API. The company reports 50% faster image generation, support for animated GIFs, and more accurate preservation of characters, products, and brand identity during editing. For marketing teams, this is primarily an upgrade to creative production tools, from rapid banner variations to animated characters for social media.
Key Takeaways
- According to OpenAI, Images 2.6 generates images 50% faster than the previous version.
- A new feature makes it possible to create animated GIFs from a text prompt or an image.
- Two models have been introduced for the API: GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst.
- Flare can generate an image in approximately six seconds at low quality settings.
- Sunburst focuses on greater detail and image consistency across successive edits.
- Images 2.6 is available globally in ChatGPT and Codex across all subscription tiers; the GIF feature in ChatGPT is rolling out gradually.
What Has Changed Since Images 2.5
OpenAI’s previous major image-generator update arrived in September 2026. Images 2.5 introduced more precise editing of specific image areas, the ability to work with sketches, and ready-made templates for creating visuals.
UAGEEK covered these capabilities in its article about ChatGPT Images 2.5 and the Sketch tool.
Images 2.6 builds on that approach. OpenAI focused on three objectives: reducing generation time, preserving details more effectively between edits, and adding animation.
This matters in workflows where a designer or media buyer does not create a single image from scratch but instead makes successive changes to an existing layout: its background, composition, objects, text elements, and format.
Image Generation Is Now 50% Faster
OpenAI says Images 2.6 generates images 50% faster than the previous generation. This is a performance claim from the developer, not a guarantee of the same speed improvement for every type of prompt.
Actual generation time depends on the selected model, quality settings, scene complexity, and conditions of service use.
For ad production, shorter waiting times can be especially useful when testing multiple versions of a creative. For example, a team might compare several compositions, backgrounds, or visual approaches for the same product.
However, faster generation does not automatically reduce total production time by 50%. Checking text, logos, brand-guideline compliance, and advertising policies remains a separate step.
GPT-Image-2.6 Flare vs. Sunburst: What’s the Difference?
Through the API, OpenAI offers two models with different priorities.
Flare — for Fast Iterations
GPT-Image-2.6 Flare prioritizes speed. According to OpenAI, it can produce an image in approximately six seconds with the quality setting set to low.
The company also reports improvements in the composition of posters, product images, and other professional visuals.
Flare can be a convenient option for early concept exploration, composition testing, and rapidly producing drafts. But six seconds represents the best-case scenario cited by the company, not a fixed duration for every request.
Sunburst — for Detail and Iterative Editing
GPT-Image-2.6 Sunburst is designed for tasks where precision matters more. OpenAI recommends using it with medium or high quality settings for better detail and consistency between edits.
For example, when changing a background, the model should better preserve a character’s clothing, a product’s appearance, or other elements that are not supposed to be edited.
This does not mean objects will be perfectly identical in every generation. Results still need to be checked for advertising materials where exact wording, packaging, or logos are important.
Animated GIFs Can Now Be Created in ChatGPT
One of the most noticeable new features of Images 2.6 is animated GIF generation.
Users can describe the animation they want in text or upload a source image. The system creates a sequence of frames and assembles them into a GIF.
OpenAI demonstrates character animations, reactions to events, brand mascot movements, and short looping scenes. ChatGPT includes options to adjust playback speed and edit individual frames.
This extends the generator’s capabilities beyond static images, but it does not make GIFs a full replacement for video generation. The format is primarily suited to short repeating movements, reactions, and simple animated elements.
For advertising campaigns, a GIF can be useful as an intermediate asset, but support must be checked separately for each ad placement. Not every platform accepts GIFs as finished advertising files.
Preserving Characters and Brands Has Become a Greater Priority
One of the main improvements in Images 2.6 is more consistent reproduction of details during repeated editing.
The problem is familiar to marketing teams: you want to change a background or camera angle, but the generator also changes the character’s appearance, the packaging’s shape, or elements of the clothing.
OpenAI says the new version preserves such details more effectively. This could make it easier to produce a series of creatives featuring the same character or product.
ChatGPT has also introduced saved reference images. These let users reuse familiar objects or characters in later conversations without uploading the original photo again.
Still, reference images and improved consistency do not replace checks against brand guidelines. If a product’s exact shape or official visual identity is critical to a company, the final version must be compared with the original.
What’s New in the API: SVG, Markup, and Complexity
The Images 2.6 update is not limited to ChatGPT. For developers, OpenAI has added tools for more precise control over image generation.
- Image markup. Allows users to mark the areas of an image that need changing rather than describing them only in text.
- SVG export. Makes it possible to obtain results in vector format for subsequent use in design, on websites, or in printed materials.
- Complexity. A new parameter that controls how much processing the model devotes to a request.
- GIF generation. Adds the ability to generate animated results through the API.
The quality and complexity parameters serve different purposes. Quality affects the visual quality of the output, while complexity determines how much effort the model spends processing a difficult request.
OpenAI recommends a low complexity setting for most straightforward edits and medium or high settings for scenes involving many objects or constraints.
SVG export should also not be taken as a guarantee that every complex generated design will be immediately ready for professional printing or production. The structure, text, and quality of vector elements must be checked before use.
How to Use Images 2.6 for Advertising Creatives
The update’s greatest practical potential lies in producing several assets in sequence from a single visual concept.
For example, a team can first create a main product image, then prepare several background and composition variations, and finally make a short animation for social media.
To minimize unintended changes, it makes sense to keep the original reference, describe the elements that must remain unchanged separately, and review each version before moving on to the next iteration.
For more complex video creatives, planning with a storyboard remains relevant. UAGEEK explained this approach in detail in its article about creating AI UGC with a storyboard.
It is important not to confuse production speed with advertising performance. Even if a model can quickly prepare dozens of variations, their CTR, CPA, and ROAS are determined by actual testing results, not by generation quality alone.
4 Billion Images a Week: What OpenAI’s Figure Means
In its release, OpenAI reported that users across ChatGPT, Codex, and the API collectively create approximately 4 billion images every week, averaging more than half a billion per day.
This is the company’s reported total generation volume across its products, not the number of images created exclusively with Images 2.6.
OpenAI did not provide a separate breakdown of advertising creatives, personal images, GIFs, or other use categories in the announcement. The figure therefore cannot be used to determine the share attributable to professional design or marketing.
Images 2.6 Availability and Pricing
OpenAI announced global availability of Images 2.6 in ChatGPT and Codex for users across all subscription tiers on the web, desktop, and mobile devices.
However, GIF creation in ChatGPT is rolling out gradually, so availability may vary between accounts.
For developers, the GPT-Image-2.6 Flare and Sunburst models are available through the API. Usage is billed at separate rates, which should be checked against OpenAI’s current pricing.
Access to the model under a particular ChatGPT plan does not mean unlimited generations. Limits and terms of use vary by product and plan.
What to Check Before Going Into Production
For marketing teams, it makes sense to evaluate Images 2.6 on their own tasks, not solely on OpenAI’s demonstrations.
- Character consistency. Does the character’s appearance remain the same after several successive edits?
- Product accuracy. Do the product’s shape, packaging, and small details remain unchanged?
- Text and logos. Does the wording match the brand’s original materials?
- Speed. How long do your typical requests actually take, rather than demonstration examples?
- Cost. What is the cost of a usable result when repeated generations are included?
- GIF. Is the animation smooth enough, and does the intended ad platform support the format?
For a team that tests many ad concepts every day, even a modest reduction in the time spent on repeated edits may be more valuable than maximizing the detail of a single image.
Conclusion
Images 2.6 makes OpenAI’s generator better suited to multi-stage visual content creation. Flare focuses on speed, Sunburst on quality and editing consistency, while GIF and SVG expand the range of available output formats.
For advertisers, the main benefit may lie not in a single fast generation but in being able to work with the same character or product through several iterations. However, final quality, brand compliance, and advertising performance still require human review.
Sources for Verification
Sources for Verification
OpenAI — Introducing Images 2.6, November 4, 2026. The official announcement covering faster generation, the Flare and Sunburst models, GIF, SVG, the complexity parameter, availability, and the statistic of 4 billion images per week.
Official Images 2.6 announcement
OpenAI — Release Notes, November 4, 2026. Confirmation of the models’ API launch, the gradual rollout of GIFs in ChatGPT, saved reference images, and new editing capabilities.
OpenAI product release history
OpenAI — ChatGPT Images 2.6 System Card. Technical description of the models, safety evaluations, limitations, and additional safeguards for GIF generation.
Images 2.6 safety document
OpenAI Developer Community — Images, upgraded, November 4, 2026. Additional GIF examples, frame editing, use of references, and an overview of the models for developers.
Developer release overview
