Persistent 429 Errors on Tier 3 Account Using Gemini 3.1 Pro Preview

Gemini API
, , , ,
1

We are experiencing persistent 429 Too Many Requests errors while using the Gemini API with the gemini-3.1-pro-preview model.

This is happening on a Tier 3 account and is significantly impacting production reliability.

Environment

  • API: Gemini API

  • Platform: Google AI Studio

  • Model: gemini-3.1-pro-preview

  • Usage Tier: Tier 3

  • Authentication: API Key

  • SDK/API Method: REST

Issue Description

We are receiving frequent HTTP 429 Too Many Requests responses, even though our usage appears to remain within expected Tier 3 limits.

Typical error:

{
  "error": {
    "code": 429,
    "message": "Resource has been exhausted",
    "status": "RESOURCE_EXHAUSTED"
  }
}

Observed Behavior

  • 429s occur repeatedly during normal production traffic

  • Exponential backoff and retries do not consistently recover

  • API reliability has become unpredictable despite Tier 3 access

Mitigations Already Attempted

  • Implemented exponential backoff with jitter

  • Reduced concurrency

  • Lowered request throughput

  • Added retry queues

Questions

  1. Are there currently any known capacity or throttling issues affecting gemini-3.1-pro-preview?

  2. Are preview models subject to additional dynamic throttling beyond documented Tier 3 quotas?

  3. Is there any hidden concurrency/session limit that could trigger 429s?

We would appreciate any clarification or guidance, as this issue is affecting production stability despite operating on a Tier 3 account.

1 Like