Skip to main content
Use the stable model ID:
Do not use the retired gemini-3-pro-image-preview ID.

Output controls

Supported ratios:
The model-specific API contract exposes the complete ratio and size matrix above. Use only those values in requests.

4K generation body

Editing body

Repeat inlineData for multiple references. Unified asynchronous editing accepts 1–16 images, with a 20 MiB decoded limit per file and 64 MiB total. Search grounding and broader tool behavior still follow the selected official model’s current support. A fixed maximum source-image byte or pixel size is not currently published. Total request-body, proxy, and upstream limits may differ; reduce the source image after HTTP 413, and contact API support before relying on a guaranteed production envelope.

Transport and logging

A 4K response can carry several megabytes of decoded image data and a larger Base64 string. Set request, proxy, memory, and log limits accordingly. Never write the full Base64 payload into normal application logs.

Run complete Pro examples

Generate, queue, edit, and handle the returned MIME.

Review exact prices

Compare Pro with Flash using current YingTu rates.
Last modified on September 3, 2026