GPT Image 2

gpt-image-2 is the best default public image model to start with. It is suitable for flow validation as well as real production use.

Its main strengths are broad coverage and richer ratio / resolution options.

Quick Facts

Item	Value
Model ID	`gpt-image-2`
Type	Image
Region	Global / CN
Best For	General image generation, posters, covers, editing
Priority	High, best default starting point

Recommended Request Shape

Image tasks still go to POST /v1/generations; only the model value changes.

{
  "model": "gpt-image-2",
  "prompt": "A premium skincare poster on warm marble, soft daylight, luxury editorial style",
  "aspect_ratio": "4:5",
  "resolution": "2K",
  "n": 1
}

Field Reference

Field	Type	Required	Default	Description
`model`	`string`	Yes	-	Must be `gpt-image-2`
`prompt`	`string`	Yes	-	Image prompt
`reference_images`	`string[]`	No	-	Reference image URL list
`aspect_ratio`	`string`	No	`auto`	Output aspect ratio
`resolution`	`string`	No	`1K`	Output resolution
`n`	`integer`	No	`1`	Image count

Supported Enums

Field	Values
`aspect_ratio`	`auto` `1:1` `5:4` `9:16` `21:9` `16:9` `3:2` `4:3` `4:5` `3:4` `2:3`
`resolution`	`1K` `2K` `4K`

`gpt-image-2` Integration Notes

If you just want to validate create-task, polling, and webhook flows, this is the best first model to test.
It supports 5:4 and 4:5, which the Banana models do not.
For lower-cost validation, start with resolution=1K.

Smallest Example

{
  "model": "gpt-image-2",
  "prompt": "A premium skincare poster on warm marble"
}