[Urgent] Critical Discrepancy: High Volume of 429 Errors despite Rate Limit Dashboard showing usage far below quota (Gemini Paid Tier)

[TECH_MIC_ACE]

Dear Google Cloud Support Team,

I am experiencing a critical issue with the Gemini API within project adsbrain-dev. We are on Paid tier 1, but our application is suffering from a very high rate of 429 TooManyRequests errors, which contradicts the data shown in the Google AI Studio dashboards.

I am attaching two screenshots from the AI Studio dashboard demonstrating this discrepancy:

1. Attachment 1 (Usage Dashboard): Shows a massive spike in “429 TooManyRequests” errors on Dec 11, reaching thousands of errors.

   

2. Attachment 2 (Rate limits breakdown): For the exact same period, the “Peak requests per minute (RPM)” chart shows our usage peaked at only ~64 RPM, which is significantly below our assigned limit of 1,000 RPM for the gemini-2.5-flash model.

   

The Problem: There is a clear contradiction between your reported usage metrics and the actual API behavior. The dashboard indicates we are well within our safe quota limits (using less than 10% of capacity), yet the API is aggressively throttling our requests with 429 errors.

Request for Investigation: Please investigate the backend logs for project adsbrain-dev to determine the true cause of these 429 errors. Specifically:

• Is there a hidden burst limit or regional quota that is not reflected in the main dashboard?

• Is it possible that the project is incorrectly being enforced under “Free Tier” limits (15 RPM) despite being configured for Paid Tier?

This issue is severely impacting our production service. We need an explanation for this discrepancy and a resolution to stop the throttling when we are within our displayed limits.

Thank you.

[MaggiR]

Hello everyone,

we are experiencing the same issue in our Tier 1 plan for a couple of days now, receiving 429 errors after just a few hundrets API calls to gemini-2.5-pro or gemini-2.5-flash each day.

What we observed (additionally to TECH_MIC_ACE’s description):

• When reached, the enforced rate limit persists for the rest of the day and seems to reset each night.
• The displayed RPD for gemini-2.5-pro and for gemini-2.5-flash is at 10K, but we receive 429 before even reaching 1K for both models (see screenshot).

We’re an academic research lab using Gemini for scientific projects. Due to approaching deadlines, we are highly depending on a quick fix for that problem, so we’d appreciate any speedy help!

[chunduriv]

Hey All,

Thank you for flagging this issue. We apologize for the inconvenience and have escalated it to our internal team for investigation. We will update you as soon as we have more information. Could you please provide the project number (not the project ID) via direct message if you have not yet done so?