Hello Google Gemini API Support,
We are experiencing repeated 503 errors when calling the image generation endpoint with our API key.
Endpoint
POST https://generativelanguage.googleapis.com/v1beta/models/gemini-3-pro-image-preview:generateContent
Model
gemini-3-pro-image-preview
Error received (example)
HTTP/1.1 503 Service Unavailable
503 UNAVAILABLE. {'error': {'code': 503, 'message': 'This model is currently experiencing high demand. Spikes in demand are usually temporary. Please try again later.', 'status': 'UNAVAILABLE'}}
Context
-
Our service generates a PNG architecture diagram (response modalities: TEXT + IMAGE).
-
We are seeing this intermittently, sometimes repeatedly, across multiple attempts and sessions.
-
We would like guidance on best practices to avoid/mitigate this (recommended retry strategy, backoff, quotas, or alternative model/endpoint).
Questions
-
Is this behavior expected for
gemini-3-pro-image-preview(capacity/preview limitation)? -
Do you recommend a different image-capable model for production workloads with better availability/SLA?
-
Are there any per-key quota/rate limits or regional constraints that could increase 503 frequency?
-
Should we implement a specific exponential backoff + Retry-After handling, and what maximum wait is recommended?
-
Can you confirm whether there is a known incident affecting this model at the moment, and any ETA/updates?
If you need additional details (request ID, project ID, headers, or timestamps), tell us what to capture and we will provide it.
Thank you,
Abid Aymen