We use gemini-3.8-flash-tts on Tier 1 through the configured endpoint generativelanguage.googleapis.com/v1beta/interactions. On November 2, 2026, between 08:06 and 08:09 UTC, 10 requests succeeded and the next request returned HTTP 429, code “too_many_requests,” stating a 100-requests-per-day limit and Retry-After: 85847 seconds.
At 08:58 UTC, after refreshing AI Studio, the project’s last-one-day view still showed RPD 10/100 and Tier 1. The response contained no quota metric or project identifier. We have not independently established the mapping between the runtime credential and the displayed project. No further generation requests were made for this investigation.
Which quota scope or accounting behavior could explain this discrepancy, and what non-secret diagnostic information would help verify it? Is there a private official channel for project-specific investigation?