GPT Image 2.5 Sunburst
GPT Image 2.5 Sunburst is the precision tier of OpenAI's GPT Image 2.5 family — the heavier sibling of Flare that spends more compute per image so text, faces and delicate edits come out exactly right. It generates from a prompt or edits up to sixteen reference images, follows instructions literally and renders exact sizes up to 4K across five quality tiers.
Input
Describe the image to generate, or the edit to apply when reference images are provided.
Optional reference images (up to 16). Leave empty to generate from text; add images to edit or combine them.
0 / 16
Output aspect ratio. Use auto to let the model pick. Ignored when an exact size is set.
Optional exact output size in pixels, including 2K and 4K. Overrides aspect ratio; sizes above 2560x1440 are experimental.
How many images to generate in one run (1-10).
Render quality from a fast low draft to max fidelity. Higher tiers are slower and cost more; auto is billed like xhigh.
Background handling. Transparent requires PNG or WebP output.
Output file format.
Compression level for JPEG and WebP outputs.
Content safety strictness. Use low for more permissive generations.
Optional identifier for your end-user, forwarded to OpenAI for abuse monitoring.
Optional OpenAI API key; when set, the call is billed to your OpenAI account by OpenAI.
Output
Run the model to see output here.
Examples
Pricing
Features
API Examples
curl -X POST \
/api/v1/model/openai/gpt-image-2.5-sunburst \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"input": {}
}'