GPT Image 2.6: Flare, Sunburst, GIFs and Markup Editing
OpenAI released GPT Image 2.6 on November 4, 2026 in ChatGPT, Codex and the API, with animated GIFs, SVG output, markup edits and a new complexity setting. It is not on Nereo yet, so the generator below runs GPT Image 2.5 Flare and Sunburst.
Text to Image
What Is GPT Image 2.6?
GPT Image 2.6 is OpenAI's latest image model family, released on November 4, 2026 as Images 2.6. It follows GPT Image 2.5 and is available in ChatGPT and Codex across all subscription tiers on desktop, mobile and web, and in the API as two models: gpt-image-2.6-flare and gpt-image-2.6-sunburst.
OpenAI says Images 2.6 generates images 50% faster than before. Flare is built for fast iteration and can return an image in as little as six seconds at low quality. Sunburst is built for detailed work and keeps images more consistent across edits at medium or high quality. New in this release are animated GIF output, static SVG output and a complexity setting, while markup-guided editing comes to the API. ChatGPT also adds reusable reference images of people, pets and characters.
GPT Image 2.6 is not available on Nereo yet. The generator on this page runs GPT Image 2.5 Flare and Sunburst, which are live on Nereo today.
What's New in GPT Image 2.6, with Official Examples
GPT Image 2.6 Markup Editing: Draw the Edit Instead of Masking It
Markup is a transparent PNG laid over the first input image. You draw circles, arrows or sketches on it to show where a change belongs, then describe the change in your prompt. OpenAI's API supports it on GPT Image 2.6 and on supported GPT Image 2.5 models, through the image edit endpoint.
| Markup | Mask |
|---|---|
A transparent PNG with annotations: arrows, circles, scribbles, color blocks or a sketch. | A black-and-white image that marks the area the model is allowed to change. |
Tells the model what you want and where, together with your prompt. | Only tells the model where; the prompt carries the instruction. |
Must match the input image's size after orientation. Cannot be combined with a mask. | The Responses API uses a mask for regional edits. |
GPT Image 2.6 Markup Example
In this example from OpenAI's image generation guide, a red circle marks one vase and GPT Image 2.6 Flare recolors it. The prompt also asks it to remove the circle from the finished photo.
| Photo with markup | Prompt | GPT Image 2.6 Flare output |
|---|---|---|
![]() | Recolor the striped vase inside the red circle to solid cobalt blue ceramic. Keep its shape, the dried stems, and the rest of the kitchen unchanged. Remove the red circle from the finished photo. | ![]() |
GPT Image 2.6 GIF Output: Animated GIFs from a Prompt or Image
In ChatGPT you can turn a prompt or image into a GIF, adjust its speed and edit individual frames. In the API, GIF output works on the generation endpoint only, not on image edits.
| API setting | Range | Default |
|---|---|---|
Frame count | 2 to 64 frames | 9 frames |
Frame duration | 20 to 2,000 ms per frame | 200 ms, for a 1.8-second loop |
Good for | Mascots, reactions, memes, product loops and animated storyboards | Sketches can be animated too |
GPT Image 2.6 GIF Examples
OpenAI's own examples: a lantern loop generated by GPT Image 2.6 Sunburst in the API, and a hand-drawn sketch animated in Images 2.6.
| Input | How it was made | Result |
|---|---|---|
A text prompt | Sunburst, 9 frames at 200 ms each. The monkey returns to its starting pose after 1.8 seconds, so the GIF loops. | ![]() |
![]() | OpenAI started from a simple pen sketch and turned it into an expressive animated character. The image shown is the animation's poster frame. | ![]() |
GPT Image 2.6 Complexity, Quality and SVG Output
Complexity sets how much the model plans before rendering. Quality sets visual detail. They are separate controls, and changing complexity has no separate fee. OpenAI recommends low complexity for most edits, because higher settings can add latency and unwanted changes.
| Control | Options | Use it for |
|---|---|---|
Complexity | Low, medium, high (default high) | Low for a background or sign change; higher for dense text and many objects |
Quality | Separate rendering setting | Low for drafts; medium or high when Sunburst must hold details across edits |
SVG output | Static vector, one per request, generation only | Icons, diagrams and illustrations for websites, apps and print |
GPT Image 2.6 Complexity Comparison
OpenAI generated these paper-cut football scenes with GPT Image 2.6 Sunburst from the same prompt at medium quality. Compare the trophy, the uniforms and the CHAMPIONS sign as complexity changes.
| Low complexity | Medium complexity | High complexity |
|---|---|---|
![]() | ![]() | ![]() |
Red team, a crown on the trophy, and the banner in the foreground. | Team crests appear on the shirts, and a laurel wreath decorates the trophy. | The uniforms change to blue paper jerseys, and the sign moves into the stands. |
GPT Image 2.6 SVG Examples
OpenAI generated these Field & Flour bakery logos from the same prompt with each model. SVG output returns editable paths, shapes and colors. We show them here as images on solid backgrounds.
| GPT Image 2.6 Flare | GPT Image 2.6 Sunburst |
|---|---|
![]() | ![]() |
How to Use GPT Image 2.6
Pick the route that fits how you work.
In ChatGPT or Codex
Images 2.6 is available across all subscription tiers. Ask for an image or GIF in the chat, then keep refining it there.
In the OpenAI API
Call the image generation or edit endpoint with gpt-image-2.6-flare or gpt-image-2.6-sunburst. OpenAI may ask you to complete API Organization Verification first.
On Nereo Today
GPT Image 2.6 is not on Nereo yet. Use GPT Image 2.5 Flare for fast drafts or Sunburst for final, detail-heavy images, with up to 16 references.
What GPT Image 2.6 Is Useful For
Client Revisions
Circle the element a client wants changed with markup and describe the change, instead of explaining its position in words.
Brand Mascots and Campaign GIFs
Animate a mascot or character into a short loop for a campaign or social post.
Reactions and Memes
Turn a prompt, photo or simple sketch into a shareable GIF reaction.
Product Imagery and Posters
Use Flare for fast layout rounds, then Sunburst when product and brand details must hold across edits.
Vector Assets
Generate an SVG icon, diagram or illustration for websites, apps and print, then check the paths before shipping.
Storyboards
Use animated storyboards to explore character movement before video production.
GPT Image 2.6 Prompt Examples
OpenAI's markup prompt, plus editing prompts that work with GPT Image 2.5 on Nereo today.
Markup Recolor (OpenAI Example)
“Recolor the striped vase inside the red circle to solid cobalt blue ceramic. Keep its shape, the dried stems, and the rest of the kitchen unchanged. Remove the red circle from the finished photo.”
From OpenAI's guide. It needs a markup layer in the API, which Nereo's editor does not have.
Recolor One Object
“Change only the mug on the left to matte cobalt blue. Keep the table, lighting, shadows and everything else exactly the same.”
Name the object and its position.
Replace Poster Text
“Replace the headline with the exact text "SUMMER SALE 40% OFF" in the same font style and position. Do not change the photo or layout.”
Use Sunburst for text-heavy edits.
Add an Element
“Add a small potted olive tree to the empty corner on the right, matching the room's light and perspective. Keep the furniture unchanged.”
Describe where it goes and what must stay.
Keep a Character Consistent
“Using the woman in reference image 1, place her in a rainy Tokyo street at night holding a clear umbrella. Keep her face, hair and jacket the same.”
Add two or three references for stronger identity.
GPT Image 2.6 vs GPT Image 2.5
| Detail | GPT Image 2.6 (Flare, Sunburst) | GPT Image 2.5 Flare | GPT Image 2.5 Sunburst |
|---|---|---|---|
| Available on Nereo | |||
| Released | November 4, 2026 | September 2026 | September 2026 |
| Best for | Flare: fast iteration. Sunburst: detailed work and consistent edits | Fast everyday images and edits | Polished output, text and tight control |
| Animated GIF output | Yes, generation only | No | No |
| SVG output | Yes, generation only | No | No |
| Complexity setting | Low, medium, high | No | No |
| Markup editing | In OpenAI's API | In OpenAI's API, not in Nereo's editor | In OpenAI's API, not in Nereo's editor |
| API price per 1M tokens (text in / image in / image out) | $5 / $8 / $30 | $5 / $8 / $30 | $5 / $8 / $30 |
| Output on Nereo | Not available | 1K, 2K, 4K | 1K, 2K, 4K |
| Reference images on Nereo | Not available | Up to 16 | Up to 16 |
GPT Image 2.6 Price
GPT Image 2.6 costs the same as GPT Image 2.5 in OpenAI's API, and Flare and Sunburst share one rate. These are OpenAI's prices, not Nereo prices.
GPT Image 2.6 API
$5 per 1M text input tokens ($1.25 cached), $8 per 1M image input tokens ($2 cached) and $30 per 1M image output tokens.
ChatGPT and Codex
Images 2.6 is included across all ChatGPT and Codex subscription tiers. Check your plan for its limits.
GPT Image 2.5 on Nereo
Generation on Nereo uses credits, and the cost of each image is shown before you generate. Any trial credits depend on your account; unlimited free generation is not included.
GPT Image 2.6 Alternatives on Nereo
GPT Image 2.5
OpenAI's current image model on Nereo, in Flare and Sunburst versions.
Nano Banana 2.1
Google's fast image generation and editing model with strong prompt adherence.

Seedream 5.0 Pro
ByteDance Seed's flagship for realism, precise text and annotation-guided edits.
GPT Image 2.6 FAQ
Find answers to common questions about our AI video generation service.
Yes. OpenAI released Images 2.6 on November 4, 2026 in ChatGPT, Codex and the API.
Flare is for fast iteration and can generate an image in as little as six seconds at low quality. Sunburst is for detailed work and keeps images more consistent across edits at medium or high quality. Both cost the same.
GIF output, SVG output and the complexity setting are new, and OpenAI says generation is 50% faster than before. Prices are unchanged. Markup editing is not exclusive to 2.6: OpenAI's API also supports it on GPT Image 2.5 models.
Use gpt-image-2.6-flare or gpt-image-2.6-sunburst with the image generation and image edit endpoints. There is no model called just gpt-image-2.6.
In OpenAI's API, $5 per 1M text input tokens, $8 per 1M image input tokens and $30 per 1M image output tokens, the same for Flare and Sunburst and the same as GPT Image 2.5.
OpenAI says Images 2.6 is available in ChatGPT and Codex across all subscription tiers, so check the limits on your plan. The API is paid. GPT Image 2.5 on Nereo uses credits, and the cost is shown before you generate.
Ask ChatGPT to turn a prompt or image into a GIF, then adjust the speed or edit individual frames. In the API, request GIF output on the generation endpoint and set 2 to 64 frames.
Markup is a transparent PNG laid over the image you are editing. You draw circles, arrows or sketches on it to show where a change belongs and describe the change in your prompt.
Yes, in the API. Set output_format to svg on the generation endpoint. It returns one static vector at a time with editable paths, shapes and colors, and it does not work on the edit endpoint.
No. A mask marks the area the model may change. Markup is an annotation layer that shows what you want and where. The two cannot be combined in one request, and markup does not guarantee that every unmarked pixel stays the same.
It controls how much the model plans before rendering: low, medium or high, with high as the default. OpenAI recommends low for most edits and medium or high for scenes with many elements. It has no separate fee.
OpenAI's guide lists custom sizes up to 3840 pixels on an edge for its GPT Image 2.5 models and calls sizes above 2560x1440 experimental. It has not published a separate limit for 2.6, so treat 4K on 2.6 as unconfirmed.
No. MAI-Image-2.6 is a separate image model from Microsoft AI.
Not yet. Nereo offers GPT Image 2.5 Flare and Sunburst today, and this page will be updated if GPT Image 2.6 is added.
No. Use your Nereo account and Nereo credits. A ChatGPT subscription does not include Nereo credits.









