Skip to main content
gpt-image-2-vip is the route with a fully exercised low, medium, high, and auto workflow. Use it when predictable access to the documented quality enum matters more than the lower Web price. This is a YingTu model ID in the GPT Image 2 family. OpenAI’s official model page publishes the gpt-image-2 alias and the dated snapshot gpt-image-2-2026-04-21; do not send the -vip ID to OpenAI directly. The current YingTu price is USD 0.05 per request. This is a fixed gateway price, not OpenAI’s token pricing; use the model marketplace as the current source.

​
Quality and size controls

VIP exposes four quality values and three documented dimensions:
Quality changes the output image token budget. It does not guarantee that a higher tier will always finish faster, and it is not an objective score for every prompt.

​
Available workflows

The model supports synchronous Generate and one-source-image Edit plus unified task-ID generation and editing with 1–16 reference images through /v1/images/tasks. Both synchronous and task results use the native ImagesResponse data shape and return one Base64 PNG. Choose gpt-image-2-web when USD 0.03 fixed pricing matters more than explicit quality control.

Review VIP parameters

Choose quality, size, and error-safe values.

Run VIP examples

Generate, queue, edit, and decode the image.
Last modified on September 3, 2026