GPT Image 2.5
OpenAI visual model with two rendering styles, an optional transparent background, and up to 4K output.
| Property | Value |
|---|---|
modelSlug | gpt-image-2-5 |
| Kind | Image |
| Maker | OpenAI |
Copies per request (variations) | 1 to 5 |
| How the cost is counted | per copy, at a rate that depends on style and resolution |
| Over MCP | yes, with the generate tool |
Credit amounts are left out because they change. The lowest price is creditCost on GET /v1/models, and POST /v1/jobs answers with the exact creditCost of a request.
input fields
Send the fields below inside the input object. A field you leave out takes its Default. Unknown field names are dropped without an error, so check the spelling if a setting seems to have no effect.
| Field | Type | Required | Default | Rules |
|---|---|---|---|---|
prompt | string | yes | – | Up to 20,000 characters |
reference_images | string[] (image upload ids) | – | – | Reference images. Up to 16 files |
style | string (one of) | – | sunburst | sunburst (Sunburst), flare (Flare) |
aspect_ratio | string (one of) | – | auto | auto, 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16, 21:9, 27:16, 16:27, 9:8, 8:9 |
resolution | string (one of) | – | 1K | 1K, 2K, 4K |
background | string (one of) | – | auto | auto, opaque, transparent |
Rules by setting
This model runs a different version depending on what you send. The rules below are checked before any credits move; a request that breaks one is refused with 400 and a message naming the field.
Every request
- Refused combinations:
resolution=2Kwithaspect_ratio=27:16;resolution=2Kwithaspect_ratio=16:27;resolution=2Kwithaspect_ratio=9:8;resolution=2Kwithaspect_ratio=8:9;resolution=2Kwithbackground=transparent;resolution=2Kwithbackground=opaque;resolution=4Kwithaspect_ratio=27:16;resolution=4Kwithaspect_ratio=16:27;resolution=4Kwithaspect_ratio=9:8;resolution=4Kwithaspect_ratio=8:9;resolution=4Kwithbackground=transparent;resolution=4Kwithbackground=opaque.
style = sunburst and without a reference file
- Takes no files in
reference_images; a request that fills one is refused.
style = sunburst and with a reference file
reference_images: required; up to 16 files.
style = flare and without a reference file
- Takes no files in
reference_images; a request that fills one is refused.
style = flare and with a reference file
reference_images: required; up to 16 files.
Example request
{
"modelSlug": "gpt-image-2-5",
"input": {
"prompt": "Product photo of white running shoes on a river stone, soft side light in the morning"
}
}