API returns "quota exceeded, limit: 0" while dashboard shows 2% usage - gemini-3-pro-preview & gemini-3.1-pro-preview Description:

Gemini API
,
1

I’m using the Gemini API (paid tier) to process data with gemini-3-pro-preview and gemini-3.1-pro-preview models.

The problem:
The API intermittently returns HTTP 429 with this error:

RESOURCE_EXHAUSTED: You exceeded your current quota
Quota exceeded for metric: generativelanguage.googleapis.com/generate_requests_per_model_per_day, limit: 0

Note: limit: 0 — the quota is set to zero, making the model completely unusable.

What the dashboard shows at the same time:

  • Request limit per minute: 2% usage (0.5 out of 25)

  • Input token count per minute: 0% usage (1.5 out of 1,000,000)

  • No daily limit is visible anywhere in the Quotas & System Limits page

Behavior pattern:

  • Models work fine for 1-2 hours (~200-400 requests), then suddenly get blocked with limit: 0

  • Sometimes both models block simultaneously, sometimes one at a time

  • After a few hours, models unblock on their own without any action from my side

  • The dashboard sometimes shows the models, sometimes they disappear entirely from the quota list, then reappear

  • Dashboard usage percentage never exceeds 8%, yet the API returns “quota exceeded”

My usage: ~3-4 requests/minute, well within the 25 req/min limit.

any hekp would be really appreciated

2

I also had the exact same issue the other day and just to be safe had to move back to 2.5 flash.

Not to mention their native audio api…first call normal, follow-up call is someone completely different posessed person. Hope the Gemini teams pick this up and communicate what is being done to improve these known issue rather than be on linkedin all day touting their models and products…when the reality is that these don’t work at scale or prod - at least for non-enterprise customers.

3

For what its worth I think I have exactly the same problem as you, thats why I’m here. Paid key applied and constant exceeded limit errors. If you do find a solution or the contact to fix this, please post it here.