AI image model · OpenAI

GPT Image 2.6, Explained: Sunburst, Flare and API

Q: Can gpt-image-2.6 make GIFs and SVGs?

Yes. Set output_format to gif with 2 to 16 frames and 20 to 2,000 ms per frame, or to svg for vector graphics with editable paths. SVG works only on the generations endpoint with one image per request, and neither format supports streaming, partial images or a custom compression setting.

GPT Image 2.6 is OpenAI's newest image model, released on November 4, 2026; this page covers its two variants, specs and API pricing. Want product photos and edits at a flat price instead? Supavisual's own generator charges 1 credit per image, about 2.4 cents on Hobbyist.

Supavisual: 3 free trial credits
1 credit per image, any tool
Commercial license on paid plans

Generate an image

2: Variants in this release: Sunburst and Flare
$30: Per 1M image output tokens on OpenAI's API
2–16: Frames per animated GIF through the API
1 credit: Per image on Supavisual's own generator

What is GPT Image 2.6?

GPT Image 2.6 is OpenAI's image generation and editing model, released on November 4, 2026 as the successor to GPT Image 2.5. In ChatGPT it ships as Images 2.6 (OpenAI's system card calls it ChatGPT Images 2.6), and OpenAI made it available globally in ChatGPT and Codex across all subscription tiers; press coverage citing OpenAI's help docs says GIF creation is rolling out in ChatGPT.

For developers, the release comes as two models with the same price: Sunburst, aimed at detailed visual work where editing precision matters most, and Flare, OpenAI's fastest model for everyday generation. Both take text and image inputs and return images through the Image API (generations and edits) or as a tool inside the Responses API.

OpenAI's guide lists four API features for both 2.6 models: a complexity setting that controls how much the model plans before rendering, markup-guided edits (which the guide also lists for 2.5), SVG output and animated GIF output. OpenAI advertises the update as about 50% faster than the previous version, with Flare returning a low-quality draft in as little as six seconds. Those speed figures are OpenAI's own; we have not measured them.

GPT Image 2.6 Sunburst vs Flare: which variant to pick

Both variants share features and token prices. The difference is what OpenAI tuned each one for.

	Sunburst	Flare
API model id	`gpt-image-2.6-sunburst`	`gpt-image-2.6-flare`
Tuned for	Detailed work, precise edits, consistency	Fast, high-quality everyday images
Quality levels	low, medium, high, xhigh, max	low, medium, high, xhigh, max
Complexity	low, medium, high (default high)	low, medium, high (default high)
SVG and GIF output	Yes	Yes
Image output price	$30 per 1M tokens	$30 per 1M tokens

Equal token prices do not mean equal cost per image: OpenAI notes that token use differs by model and quality setting.

GPT Image 2.6 specs: inputs, outputs and API limits

These are OpenAI's documented limits as of November 2026. The last row shows Supavisual's own generator for comparison.

Maker	OpenAI
Released	November 4, 2026
Inputs	Text prompts and input images (edits endpoint, Responses API)
Raster output	PNG by default, or JPEG and WebP with adjustable compression
SVG output	Generations endpoint only, one image per request (`n` = 1)
GIF output	2–16 frames (default 9), 20–2,000 ms per frame (default 200 ms)
Markup edits	Transparent PNG annotation layer, same size as the first image, up to 50 MB
Access	ChatGPT, Codex, Image API, Responses API; hosted only, no downloadable weights
Alternative: Supavisual's generator	1–4 images per prompt, up to 6 reference images, 5 aspect ratios, 1 credit each

What's new in GPT Image 2.6

OpenAI's guide lists the first four for both Sunburst and Flare in 2.6; the last two carry over from version 2.5.

Complexity control

A complexity field (low, medium or high) sets how much the model plans before it renders. Low suits a focused edit such as swapping a background; high helps with dense text and busy layouts. It has no separate fee.

Markup-guided edits

Draw circles, arrows or sketches on a transparent PNG layer and send it with the photo. The model edits the marked areas. Markup applies to the first input image and can't be combined with a mask.

SVG output

Set output_format to svg for icons, logos and flat illustrations with editable paths. The SVG is static vector content: no scripts, external resources or embedded bitmaps.

Animated GIFs

Set output_format to gif and describe the motion and loop. You choose 2 to 16 frames and the duration of each one, from either the generations or the edits endpoint.

Five quality levels

Quality runs from low to max. Low is the cheap draft setting; xhigh and max cost more output tokens for final assets.

Multi-turn editing

Through the Responses API you can keep refining the same picture over several turns, much like editing a picture in ChatGPT.

GPT Image 2.6 pricing (as of November 2026)

OpenAI bills the model by tokens, not per picture. Sunburst and Flare cost the same.

Token type	Standard, per 1M tokens	Cached input, per 1M tokens
Image input	$8.00	$2.00
Image output	$30.00	Not applicable
Text input	$5.00	$1.25
Batch processing	50% off standard rates	Not applicable
Supavisual, for comparison	$19 a month for 800 images (1 credit each)	First month $9 on Hobbyist

OpenAI prices as of November 2026. Your cost per image depends on size, quality level and how many input images you send.

How to use the GPT Image 2.6 API, and where else it runs

There are three official ways to reach GPT Image 2.6. The simplest is ChatGPT: every subscription tier gets ChatGPT Images 2.6 in the chat window, on the web, desktop and mobile. Codex has it too, for image work inside coding tasks.

For the GPT Image 2.6 API, set model to gpt-image-2.6-sunburst or gpt-image-2.6-flare on the Image API, or name one of them in the image generation tool of the Responses API. The Image API is the better fit for one prompt and one result; the Responses API suits multi-turn editing. OpenAI may ask your organization to complete API Organization Verification before you can call GPT Image models, and it recommends sending a hashed safety_identifier for apps with many end users.

Keep in mind that SVG and GIF responses don't support streaming, partial images or a custom output_compression. If you need a pixel-perfect product shot instead, look at the prompt editor or a dedicated product studio.

GPT Image 2.6 limits and how to work around them

Most failed Sunburst and Flare requests trace back to a handful of documented rules. Check these before you debug your prompt.

Do

Use quality: low for drafts, then rerun the keeper at a higher level
Match the markup layer's width and height to the first input image
Set n to 1 when you ask for SVG output
Lower frame_count or frame size if a GIF request exceeds the pixel budget

Don't

Send an SVG as a raster input to the edits endpoint; it isn't supported
Combine markup with mask in one request
Expect streaming or partial images for GIF and SVG
Assume equal token prices mean equal per-image cost across variants

A GPT Image 2.6 alternative: Supavisual's generator

Supavisual does not run OpenAI's model; it is a separate image tool built for product and marketing visuals, priced per image instead of per token. In the generator you type a prompt, attach up to 6 reference images, choose one of 5 aspect ratios (1:1, 3:2, 2:3, 16:9, 9:16) and get 1 to 4 images. Every image action costs 1 credit, whatever the tool.

That flat price is the main difference from a token-billed API. Hobbyist is $19 a month for 800 credits, about 2.4 cents per image, and the first month is $9. Unlimited is $99 per seat per month, at full speed for 1,000 credits per seat, then queued. Both plans include a commercial license.

The same credit also covers editing: background removal, upscaling, expanding, retouching and smart resize into 8 social presets or a custom size. You can translate the text in an image into 37 languages, or batch edit up to 10,000 images with one prompt. New accounts get 3 free trial credits, with watermarked previews.

GPT Image 2.6 vs Nano Banana 2.1, Seedream 5.5 and Ideogram 4.0

A quick comparison of the image models people search for most this month. Each has its own page.

	GPT Image 2.6	Nano Banana 2.1	Seedream 5.5	Ideogram 4.0
Made by	OpenAI	Google	ByteDance	Ideogram
Released	Nov 2026	Oct 2026	Nov 2026 (on reseller APIs)	Jun 2026
Weights	Hosted only	Hosted only	Hosted only	Open (non-commercial)
Stand-out feature	SVG and GIF output	Up to 14 reference images	Transparent RGBA output	Text and bounding-box layout
Published API price	$30 / 1M output tokens	$0.0336 per 1K image	Not listed by ByteDance	$0.03–$0.10 per image

Prices as of November 2026. Read more: Nano Banana 2.1, Seedream 5.5 and Ideogram 4.0.

More models and Supavisual tools

Other model explainers, plus the Supavisual tools that turn a prompt or photo into a finished visual.

Frequently asked questions

What is GPT Image 2.6?

GPT Image 2.6 is OpenAI's image generation and editing model, released on November 4, 2026. It comes in two variants, Sunburst for detailed, precise work and Flare for fast everyday images, and adds complexity control, markup-guided edits, SVG output and animated GIF output. In ChatGPT it appears as Images 2.6, which OpenAI's system card calls ChatGPT Images 2.6.

How much does the GPT Image 2.6 API cost?

As of November 2026, OpenAI charges $8 per million image input tokens, $30 per million image output tokens and $5 per million text input tokens, with cached inputs cheaper and batch jobs at half price. Sunburst and Flare cost the same per token, but the cost of one image still depends on its size and quality level.

Is GPT Image 2.6 free?

OpenAI made ChatGPT Images 2.6 available across all ChatGPT subscription tiers, so you can try it in ChatGPT without a separate fee beyond your plan. The API is paid by the token. For a different free start, Supavisual gives every new account 3 trial credits for its own image tools, with watermarked previews.

What is the difference between Sunburst and Flare?

Both belong to the 2.6 release and have the same features and token prices. Sunburst is tuned for detailed visual work where edit precision and consistency matter. Flare is OpenAI's fastest image model for everyday generation, and OpenAI says it can return a low-quality draft in about six seconds. Pick Sunburst for final assets, Flare for quick iterations.

Can gpt-image-2.6 make GIFs and SVGs?

Yes. Set output_format to gif with 2 to 16 frames and 20 to 2,000 ms per frame, or to svg for vector graphics with editable paths. SVG works only on the generations endpoint with one image per request, and neither format supports streaming, partial images or a custom compression setting.

Does Supavisual use GPT Image 2.6?

No. Supavisual is a separate image tool with its own generator and does not offer GPT Image 2.6. It is built for product and marketing visuals: generate from a prompt and up to 6 references, then edit, remove backgrounds, upscale, resize and translate, all at 1 credit per image. Hobbyist is $19 a month for 800 credits.

Do I need organization verification for the OpenAI image API?

Possibly. OpenAI says you may need to complete API Organization Verification in your developer console before calling GPT Image models. It also recommends a stable, hashed safety_identifier on each request for apps that serve end users; the older user parameter is deprecated but still accepted for existing requests.

Is GPT Image 2.6 better than Nano Banana 2.1?

It depends on the job. OpenAI's model is the one to pick for SVG logos, animated GIFs and markup-guided edits. Nano Banana 2.1 accepts up to 14 reference images and lists a fixed price per image at 1K, 2K and 4K. See Nano Banana 2.1 for its full specs and pricing.

Try a flat-price GPT Image 2.6 alternative

Read up on OpenAI's model, then see what Supavisual makes from your own prompt or product photo: 1 credit per image, 3 free trial credits to start.

Generate an image

More ai image models: see all.