OpenAI Images Compatibility

This set of models uses a unified image entry route and is compatible with the OpenAI Images style externally; the actual model is determined by the model field.

Supported Models

Model Name	Description	Generation	Editing	Default Quality	Default Size
`gpt-image-1`	The primary model explicitly handled in the current unified image entrypoint	Supported	Supported	`auto`	Determined by upstream
`dall-e-3`	DALL·E 3 compatible model	Supported	Depends on the channel	`standard`	`1024x1024`
`dall-e-2`	DALL·E 2 compatible model	Supported	Depends on the channel	No enforced default	`1024x1024`
`dall-e`	Legacy DALL·E alias	Supported	Depends on the channel	No enforced default	`1024x1024`

Supported Capabilities

Capability	Entrypoint	Description
Text-to-image	`/v1/images/generations`	Standard generation entrypoint
Image editing	`/v1/images/edits`	Editing entrypoint
Image editing alias	`/v1/edits`	Legacy alias
Image variations	`/v1/images/variations`	Route retained, currently not implemented

Supported Modes

Mode	Detection Condition	Entrypoint
Text-to-image	Only `prompt` is provided, with no editing material	`/v1/images/generations`
Image editing	`image` or multi-image references are provided, together with `prompt`	`/v1/images/edits` or `/v1/edits`
Inpainting	`mask` is additionally provided in the editing request	`/v1/images/edits` or `/v1/edits`

Output Constraints

Field	Constraint
`response_format`	Common values are `url` and `b64_json`
`n`	When omitted or explicitly set to `0`, the unified layer falls back to `1`
`size`	`dall-e-2` only allows `256x256`, `512x512`, and `1024x1024`
`size`	`dall-e-3` only allows `1024x1024`, `1024x1792`, and `1792x1024`

Unified Entrypoint Notes

The unified layer first parses the request in the OpenAI Images style, then dispatches it to the specific image channel.
model is a required field; the unified layer will first validate whether it exists.
watermark is carried as a pointer field; explicitly passing false is semantically different from omitting it.
When requesting b64_json or base64, the response will use Base64 image data instead.

Subpage Entries

Image Generation

POST /v1/images/generations

Image Editing

POST /v1/images/edits and POST /v1/edits

Recommended Reading Order

Midjourney Task Query OpenAI Images Compatible Image Generation