Critical Discrepancy: High 429 Errors despite being <10% of Paid Tier Quota (Project: aigrademypaper))

Google AI Studio
, , ,
1

Hi everyone,

I’m reaching out to see if anyone else is experiencing a mismatch between their Google AI Studio Dashboard and actual API behavior. We are currently on Paid Tier 1 for the project:aigrademypaper, but our application is being crippled by 429 TooManyRequests errors that don’t align with our quota.

The Discrepancy:

  • Observed Behavior: Massive spikes in 429 errors (lots of failures) starting around May 4.

  • Dashboard Metrics: Our “Peak requests per minute (RPM)” chart shows we peaked at 31 RPM.

  • Assigned Limit: Our limit for gemini-3-flash is 1,000 RPM.

Evidence: I’ve attached two screenshots from the AI Studio dashboard showing the contradiction:

  1. Usage Dashboard: Shows the high volume of 429 errors.

  2. Rate Limits Breakdown: Shows our peak RPM is significantly below the 1,000 RPM threshold (using less than 10% of our capacity).

    Rate_Limit
    Rate_Limit1680×399 29.1 KB

Questions for the Community/Support:

  1. Is it possible the project is being throttled under Free Tier limits despite being configured for the Paid Tier?

  2. Are there “hidden” burst limits or per-IP/regional constraints that aren’t surfaced in the standard AI Studio charts?

  3. Has anyone else noticed the “Usage” dashboard reporting errors that the “Quota” dashboard says shouldn’t exist?

This is severely impacting our production environment. If any Google team members could look into the backend logs for aigrademypaper to see why the enforcement doesn’t match the dashboard limits, it would be greatly appreciated.