GPT Image 2 API routes - YingTu API Docs

GPT Image 2 is OpenAI’s current image generation and editing model. Its official model ID is gpt-image-2, and OpenAI currently lists the dated snapshot gpt-image-2-2026-04-21. The official model page identifies text and image inputs, image output, and the Image API Generate and Edit endpoints. YingTu exposes two separate public route IDs for this model family. These suffix IDs belong to this API and should not be sent to OpenAI directly.

Choose a route

Public model ID	Current price	Choose it when
`gpt-image-2-web`	USD 0.03/request	You want the lower fixed-price reverse-image route. `quality` is forwarded, with upstream-dependent effect.
`gpt-image-2-vip`	USD 0.05/request	You want the route whose four quality values have been exercised across the documented workflow.

Both routes use the same three documented output presets:

1024x1024   2048x2048   3840x2160

The final preset is the documented 4K landscape option. Do not substitute 4096x4096: OpenAI’s image generation guide limits the longest edge to 3840 pixels and the output area to 8,294,400 pixels.

Supported REST workflows

Generate: POST /v1/images/generations with JSON.
Edit: POST /v1/images/edits with a multipart source image and prompt.
Task-based generation or editing: POST /v1/images/tasks with action=generate|edit, then GET /v1/tasks/{task_id}.

The task endpoint is a YingTu extension for clients that cannot keep one connection open. It is not an official OpenAI Images endpoint. Use /v1/images/tasks for asynchronous generation or editing with 1–16 reference images. Synchronous requests return one Base64 PNG in data[0].b64_json. Unified tasks preserve the native ImagesResponse under result, so decode result.data[0].b64_json; no gateway-specific MIME field is added.

What is outside this contract

The maintained reverse-image subset supports multiple reference images but returns one output. It does not promise masks, partial-image streaming, transparent backgrounds, output-format controls, callbacks, cancellation, or OpenAI SDK compatibility. Use the documented REST requests even when an upstream guide describes a broader surface.

Compare Web and VIP

Choose by quality control and current fixed price.

Start with Web

Use the lower fixed-price route with the documented image controls.

Start with VIP

Select an explicit quality tier.

View current prices

Recheck pricing before production use.

Review SDK support

Use the maintained REST path and understand current SDK limits.

Contact API support

Send a redacted request or task identifier when documented recovery fails.

​Choose a route

​Supported REST workflows

​What is outside this contract