Skip to main content
YingTu bills the current Nano Banana image models per request at the selected output tier. The exact amount depends on the model, image size, account, and group visible when the request is sent. Use the dated YingTu pricing table for current numbers. Google prices, subscriptions, free tiers, and other API providers do not apply to the YingTu account described here.

​
One request is one output

The documented image flow accepts one candidate:
  • omit generationConfig.candidateCount, or set it to 1;
  • do not send a top-level n;
  • send separate requests for separate images;
  • each separate request is billed independently.
Synchronous and asynchronous methods use the same model price and parameter validation. Choosing async does not add a second generation charge.

​
Asynchronous reservation and settlement

Add an Idempotency-Key before submitting work that may be retried. The same key and normalized request return the original task; a changed request returns HTTP 409.
This is an adjustment to API account quota, not a cash refund. No fixed completion-time commitment is currently published for that adjustment. If a task is terminally failed and the expected quota has not returned, keep the task_id and contact API support.

​
Keep cost decisions reproducible

Store the model ID and requested size with your own job record. Compare models against the same acceptance criteria and retry budget rather than choosing only by the lowest request price.

Review exact prices

Check the current YingTu-only table and live pricing endpoint.

Submit asynchronously

Understand HTTP 202, task ownership, terminal states, and retrieval.

Retry without duplicate work

Separate safe retries from ambiguous task creation.

Review data handling

Understand task data, metadata, and the current retention boundary.
Last modified on September 3, 2026