GPT Image 2.6 is live: images in as little as 8 seconds, animated GIFs and SVG vectors from one request, official OpenAI pricing

GPT Image 2.6, OpenAI’s newest image model, is live on LaoZhang API in Flare and Sunburst variants, through official forwarding and as a per-image VIP version. One request returns a looping GIF or an editable SVG, and Flare renders a 1024 image in 8–12 seconds. Token prices match OpenAI and GPT Image 2.5.

Published: November 5, 2026
Last verified: November 5, 2026
Status: Active

At a glance

Item	Details
Models	`gpt-image-2.6-flare` (fast), `gpt-image-2.6-sunburst` (more detail, steadier edits)
Endpoints	Generate with `POST /v1/images/generations`, edit with `POST /v1/images/edits`
Token	Group `OpenAI Official` or `OpenAI Official Enterprise`; billing mode Usage first or Usage-based
Price	Per token: text input $5, image input $8, image output $30 per 1M tokens, the same as OpenAI and GPT Image 2.5; the Enterprise group adds 20%
Speed	1024×1024: Flare 8–12 s (`low` to `high`), Sunburst about 35–60 s
New	`output_format: "gif"` returns a looping animation; `output_format: "svg"` returns an SVG vector image (enabled the afternoon of November 5); `complexity` controls scene complexity at no extra cost
Per-call routes	`gpt-image-2.6-flare-vip` and `gpt-image-2.6-sunburst-vip`: $0.03 per image in the `default` group (live the evening of November 5), `quality` up to `high`, no GIF or SVG; the web model is still 2.5

Times are single runs against OpenAI on November 5 and vary with queueing. The model and pricing list is the price of record.

GPT Image 2.6 Flare render of a rainy Taipei night market with sharp small text on neon signs and reflections on the wet street — gpt-image-2.6-flare, 1536×1024, high, 12 s

Looping GIF from GPT Image 2.6 Flare: an orange cat stretching and yawning on a rainy windowsill — GIF returned by gpt-image-2.6-flare: 512×512, 9 frames, about 11 s

What you can build

Use case	Request	Result
Stickers, reactions, small product loops	`"output_format": "gif"`, `"size": "512x512"`, `"frame_count": 9`	A looping GIF in about 11 s for about $0.06
Interactive apps that need fast images	Flare with `"quality": "low"` or `"high"`	1024 images in 8–12 s
Posters, signage and packaging with small text	Flare or Sunburst with `"quality": "high"`	Legible text at about $0.053 per 1024 image
Icons and logos as vectors	`"output_format": "svg"`, `"quality": "medium"`	An editable SVG that scales cleanly, about $0.06 and 40 s for a simple icon
Retouching and targeted edits	Sunburst on `/v1/images/edits`	Keeps layout and lighting, changes only what you ask
4K banners	`"size": "3840x2160"`, `"quality": "high"`	Native 4K for about $0.10; Flare takes about 21 s

How to use it

If you already call GPT Image 2.5 through an official-forward group, change only model:

{ "model": "gpt-image-2.6-flare", "quality": "high" }

Full request for a GIF:

export LAOZHANG_API_KEY="your-official-forward-group-key"

curl https://api.laozhang.ai/v1/images/generations \
  --max-time 300 \
  -H "Authorization: Bearer $LAOZHANG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2.6-flare",
    "prompt": "An orange cat stretches on a windowsill, then yawns while light rain falls outside, flat illustration style, looping animation",
    "output_format": "gif",
    "size": "512x512",
    "frame_count": 9,
    "frame_duration_ms": 200,
    "quality": "medium"
  }'

Decode data[0].b64_json and save it as a .gif file. GIF notes:

frame_count sets the number of frames (default 9). frame_duration_ms sets each frame’s duration in milliseconds (default 200, minimum 20).
All frames are rendered on one sheet, so a small size with many frames fails with No supported sheet fits. 512x512 with 4–9 frames works as is.
GIF output does not support stream.

For SVG, set output_format to svg. The text model that writes the SVG defaults to gpt-6.1-terra and can be changed with svg.model; see Generate SVG vector images. Parameters, edit requests and transparent backgrounds are covered in the GPT Image 2.5 / 2.6 guide.

Cost per image by quality

Image output tokens and cost for one 1024×1024 image in the OpenAI Official group, measured on November 5. The prompt is billed as text input and usually costs under $0.001. Reference images in edits are billed as image input, about $0.008 for one 1024 image.

`quality`	2.6 output tokens	2.6 per image	2.5 per image
`low`	824	$0.025	$0.006
`medium`	1,449	$0.043	$0.013
`high`	1,756	$0.053	$0.053
`xhigh`	3,122	$0.094	$0.094
`max`	7,024	$0.211	$0.211

From high up, both generations cost the same. For high-quality output, switch to 2.6: same price, and Flare is faster (2.5 Flare high takes about 27 s).
At low and medium, 2.6 uses 3–4 times as many tokens as 2.5. For large batches of drafts or thumbnails, 2.5 is cheaper.
Landscape 1536x1024 uses fewer tokens than square: 664 at low, 1,353 at high.
Always send quality. Without it, Flare picks a level from the prompt, and a detailed prompt can use about 5,800 tokens for one 1024 image (about $0.17). Sunburst defaults to high.

Who is affected

Official-forward group users: you can switch to 2.6 with the same token. The 2.5 official-forward models stay available.
default and other per-call group users: gpt-image-2.6-flare-vip and gpt-image-2.6-sunburst-vip are new at $0.03 per image, with the same token. The 2.5 VIP models and the web model are unchanged.
Existing code that calls gpt-image-2.5-*: not affected.

FAQ

Can a default-group token call 2.6?

Yes, with the VIP model names gpt-image-2.6-flare-vip or gpt-image-2.6-sunburst-vip, at $0.03 per image:

quality goes up to high. xhigh or max returns 400 and is not charged.
Output is PNG only. GIF and SVG return 400; use official forwarding for those.

The 2.6 model names without the -vip suffix are official-forward only. Create a token in Token management with group OpenAI Official or OpenAI Official Enterprise and billing mode Usage first or Usage-based. See GPT Image 2.6 VIP per image.

Should I use 2.6 or 2.5?

Animated GIFs or SVG vectors: 2.6 only.
high quality or above: 2.6, at the same price and faster with Flare.
Large volumes of low drafts or thumbnails: 2.5 costs less.
A fixed $0.03 per image: 2.6 VIP up to high, 2.5 VIP when you need xhigh or max.

Do Flare and Sunburst cost the same?

Yes. The same size and quality use the same number of tokens. Flare is faster; Sunburst is slower and better suited to detailed scenes and careful edits.

What does complexity do?

complexity controls how complex the scene is and does not change the price:

Values: low, medium or high; the default is high.
Token usage stays the same at a given quality level.

How is SVG billed?

In two parts, both at official rates:

The reference image is billed by quality and size, like any other image.
Writing the SVG is billed at the rate of the text model used. A simple icon costs about $0.012–0.015 with the default gpt-6.1-terra and about $0.09–0.16 if you choose gpt-5.5.

More complex scenes produce longer SVG code, so cost and time go up; SVG works best for icons and simple graphics. A model comparison is in Generate SVG vector images.

Am I charged for failed requests?

No charge when the upstream returns a non-200 response, including parameter errors and moderation blocks. See Billing and failures.

OpenAI API pricing: token prices for GPT Image 2.6 Flare and Sunburst (checked November 5)
OpenAI image generation guide: GIF and SVG output, complexity and other parameters
GPT Image 2.5 / 2.6 guide: routes, groups, billing and examples
Model and pricing list, usage logs

​At a glance

​What you can build

​How to use it

​Cost per image by quality

​Who is affected

​FAQ

​Can a default-group token call 2.6?

​Should I use 2.6 or 2.5?

​Do Flare and Sunburst cost the same?

​What does complexity do?

​How is SVG billed?

​Am I charged for failed requests?

​Sources and related links