Skip to main content
YingTu exposes four Nano Banana image models through its own API account, billing, and task system. Requests use a Gemini-native body at the YingTu base URL, with one public authentication convention across the family.

​
What YingTu currently exposes

The public contract is the image-generation and editing subset implemented by this API. It is not a promise that every field in Google’s latest Gemini or Interactions APIs is available here. The maintained client contract is REST. Review SDK and protocol compatibility before configuring a Google or OpenAI SDK with a custom base URL.

​
Choose your first route

Compare the four models

Choose by lifecycle, output size, workload, and relative cost.

Generate a first image

Send a complete server-side request and save the returned file.

​
Current model IDs

  • gemini-2.5-flash-image: legacy Nano Banana; still visible on YingTu today, but Google lists an earliest shutdown date of October 2, 2026.
  • gemini-3.1-flash-lite-image: current 1K efficiency model.
  • gemini-3.1-flash-image: general-purpose 1K, 2K, and 4K model.
  • gemini-3-pro-image: premium 1K, 2K, and 4K model for complex assets.
Do not start a long-lived integration on gemini-2.5-flash-image without a migration plan. See the model comparison for the current YingTu alternatives.

​
Synchronous or asynchronous

Use synchronous generation when your client and proxy can keep one connection open for the final result. Use asynchronous submission when the work must survive the client connection: store the returned task_id, then poll the task until it reaches completed, failed, or expired. The asynchronous endpoint uses a provider-neutral JSON body rather than the synchronous provider body. Its task envelope is shared, while result remains Gemini-native for Nano Banana and Images-native for GPT Image. Use an Idempotency-Key for retryable submission and copy successful output before the 24-hour result window ends.

Set up authentication

Create a development key and keep it out of client code.

Review current prices

Use the dated YingTu price table before production traffic.

Plan production traffic

Understand 5,000 RPM access capacity, upstream 429 responses, and the availability objective.

Browse the model marketplace

Confirm current model availability, API type, and price.

Review data handling

Understand upstream transfer, training use, metadata, and retention boundaries.

Contact API support

Prepare a safe diagnostic report with request or task identifiers.
Last modified on September 3, 2026