GPT-Image-2.6 Sunburst
GPT-Image-2.6 Sunburst is OpenAI’s November 2026 image-generation and editing model for detailed compositions and precise edits. Its Image API supports text and image inputs with image output. The documented examples below have not been executed against this model.
- Images 2.6 explained
Release context, model choice, format restrictions and accurately labeled Codex workflow examples.
- GPT-Image-2.6 Flare
Fast iteration: exact API ID, formats, controls, token rates and documented requests.
- GPT-Image-2.5 Sunburst
Keep the previous version as a separate migration baseline.
- AI Model API
Understand model, endpoint, access and billing boundaries.
🚀Function Overview
OpenAI positions Sunburst for detailed compositions and precise edits. Use the exact Image API model ID and verify output-format restrictions. No direct Sunburst API request or matched performance evaluation was executed here; the linked Codex samples have an unidentified backend.
Key Features
- Model alias gpt-image-2.6-sunburst; the current card lists no separate dated snapshot.
- Text and image inputs; image output. /v1/images/generations and /v1/images/edits are supported; the 2.6 card does not list direct Responses or Chat Completions support.
- Generate PNG, JPEG or WebP; SVG and animated GIF are generation-only formats, not edit outputs.
- Complexity: low, medium or high, default high. Quality: auto, low, medium, high, xhigh or max. These are separate controls.
- Markup edits accept an aligned transparent PNG overlay matching the first source image. Omit input_fidelity for 2.6 edits.
- GIF: frame_count 2–64 (default 9); frame_duration_ms 20–2000, rounded to 10 ms (default 200). Model/request pixel limits also apply.
- SVG: n=1; static vector content. A generated reference image is used for SVG authoring; image and authoring-model tokens both contribute to cost.
- GIF and SVG exclude streaming, partial images, output_compression and prompt_rewriter. Only use GIF frame controls with GIF output.
- Use PNG or WebP for transparent backgrounds and inspect decoded alpha. The linked Codex samples do not identify their backend model.
- OpenAI positions Sunburst for detailed compositions and precise edits. This is provider positioning, not an independent quality ranking.
- Published per-million-token rates: $5 text input, $1.25 cached text input, $8 image input, $2 cached image input, $30 image output. Listed cached rates do not verify cache eligibility for a particular direct request.
Use Cases
- •Detailed product or campaign imagery
- •Reference-led edits with identity checks
- •Dense layouts with quality/complexity comparisons
- •SVG illustrations and short GIF loops
⚙️Input Parameters
model
stringSet gpt-image-2.6-sunburst explicitly at https://api.openai.com/v1/images/generations or /v1/images/edits. This is not a ChatGPT subscription model-picker label.
prompt
stringDescribe the scene or exact edit and the details that must survive. For GIFs describe the movement and loop.
image / markup
fileEdit input files; markup is an optional transparent PNG overlay matching the first image. This input is not demonstrated by the JSON generation examples.
quality / complexity
stringquality: auto, low, medium, high, xhigh, max. complexity: low, medium, high (default high). Tune separately.
size / background
stringUse a supported pixel size or auto. Background accepts auto, opaque or transparent; use PNG/WebP for raster alpha.
output_format
stringpng, jpeg or webp; generation additionally supports svg and gif. SVG requires n=1; both special formats have separate restrictions.
frame_count / frame_duration_ms
integerGIF only: 2–64 frames and 20–2000 ms per frame, rounded to 10 ms; defaults 9 and 200. Respect pixel limits.
💡Usage Examples
Example 1
Input Parameters
{
"model": "gpt-image-2.6-sunburst",
"prompt": "Create a clean ceramic mug product photograph with the exact text NORTH 26 on a transparent background.",
"size": "1536x1024",
"quality": "medium",
"complexity": "low",
"background": "transparent",
"output_format": "png"
}Output Results
Example 2
Input Parameters
{
"model": "gpt-image-2.6-sunburst",
"prompt": "Create a seamless loop of an orange toy robot waving once, with a fixed camera and consistent body.",
"output_format": "gif",
"frame_count": 9,
"frame_duration_ms": 200,
"size": "auto",
"quality": "medium",
"complexity": "low"
}Output Results
Example 3
Input Parameters
{
"model": "gpt-image-2.6-sunburst",
"prompt": "Create an editable two-color vector illustration of a cup beside a leaf, with no text.",
"output_format": "svg",
"n": 1,
"quality": "medium",
"complexity": "low"
}Output Results
Quick Actions
Technical Specifications
- Hardware Type
- OpenAI-hosted Image API; no public self-hosted checkpoint or hardware specification is asserted
- Commercial Use
- Subject to OpenAI service terms, usage policies and third-party rights. Output ownership terms are not a guarantee of uniqueness or copyright eligibility.
- Pricing
- $5/M text input; $1.25/M cached text input; $8/M image input; $2/M cached image input; $30/M image output. Per-result cost varies; SVG additionally includes authoring-model tokens.
- Platform
- OpenAI