
GPT Image 2.6 Flare API
Flare is built for speed. OpenAI reports about six seconds per image at its fastest quality level, which suits drafts and apps where someone is waiting on the result.
Call GPT Image 2.6 Flare and Sunburst from your own code, for still images and animated GIFs.

The GPT Image 2.6 API is programmatic access to OpenAI Image 2.6, the image models released on November 4, 2026. You send a prompt and settings from your own code and get an image or an animated GIF back, without opening a browser.
It suits work a person should not repeat by hand: generating images inside your own product, producing a batch from a list of prompts, or turning a mascot or a product shot into a looping GIF.
Access through this site is by request, so we can match it to what you are building.
The GPT Image 2.6 API family has two models, GPT Image 2.6 Flare and GPT Image 2.6 Sunburst. They accept the same parameters, so switching between them means changing one value.

Flare is built for speed. OpenAI reports about six seconds per image at its fastest quality level, which suits drafts and apps where someone is waiting on the result.

Sunburst is built for output quality. It suits final images, dense text and complex layouts at High and Ultra quality.
A GPT Image 2.6 request needs a prompt and a model. Every other field has a default, listed below.
A request returns a task id and a status. When the status is succeeded, the task lists the URL of every output.
Quality affects the price and complexity does not, which is worth knowing when you plan a high-volume workload. GIFs use frame_count and frame_duration_ms, and they start from a text prompt without input images.
| Field | Values | What it does |
|---|---|---|
| prompt | Text, up to 4,000 characters | Describes the image or GIF to generate |
| model | flare, sunburst | Picks the model |
| aspect_ratio | 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16. Default 1:1 | Sets the shape of the output |
| quality | low (Standard), medium (High), high (Ultra), xhigh (Extreme), max (Max), auto (Auto). Default low | Sets render detail and price. max is for still images |
| complexity | low (Simple), medium (Balanced), high (Detailed). Default high | Sets how much the model plans before it renders |
| output_format | gif, webp, png, jpeg. Default webp | Sets the file format. gif makes an animated GIF |
| frame_count | 2 to 16. Default 9 | Sets how many frames a GIF has |
| frame_duration_ms | 20 to 2000. Default 200 | Sets how long each GIF frame shows |
| background | auto, transparent, opaque. Default auto | Transparent needs png or webp |
| number_of_images | 1 to 10. Default 1 | Sets how many outputs one call returns |
| input_images | Up to 4 images, each under 5 MB | One image to edit, or several to combine. Not used with GIFs |
GPT Image 2.6 API keys on this site are issued on request. Send one short email and test your prompts in the browser in the meantime.

Email us with what you are building and roughly how many images or GIFs you need each month. We will reply with the next steps.

The generator below runs the same two models with the same quality and complexity options, so you can work out your prompts and settings before you write any code.