Quota persistently stuck - Gemini 3 Pro/Flash TPM exceeded for days despite no API usage

Gemini API
,
1

Issue Summary

I’ve been experiencing persistent “429 RESOURCE_EXHAUSTED” quota errors for several days despite not using my API key at all. The quota appears to be stuck and not resetting.

Environment Details

  • Project: Free Tier (Paid Tier 1)
  • API Keys: AI Studio
  • Status: 6 février 2026

Problem Description

Since a few days ago, my API project is throwing 429 errors for Gemini 3 Pro and Gemini 3 Flash models. The issue:

  1. I have NOT called these models via API for several days

  2. I can see in “Usage and Limits” that:

    • Gemini 3 Pro: TPM 1.23M / 1M (EXCEEDED)
    • Gemini 3 Flash: TPM 1.87M / 1M (EXCEEDED)
    • Both are on 28-day window

  3. Other models (Gemini 2.5 Flash, Flash Lite, etc.) are all under their limits with very low usage

Expected Behavior

If I haven’t used an API for several days:

  • TPM quota should reset within minutes/hours (sliding window)
  • RPD should reset at midnight PT
  • The quota error should clear

Actual Behavior

  • Still getting 429 errors
  • TPM counters still show exceeded even though no recent usage
  • This looks like a quota reset bug

Workaround

I’m currently switching to Gemini 2.5 Flash models which have plenty of quota remaining, but I’d like to understand if this is a known issue or if my project is in a stuck state.

Questions

  1. Is there a known issue with TPM quotas not resetting on Gemini 3 Pro/Flash?
  2. Should I create a new project to get unstuck?
  3. Is there a manual reset option available?

Thanks for any insights on this!

1 Like