// WHAT IT MEANS
You exceeded the per-minute request ceiling — 60 requests per minute on the free tier. It is a burst limit, separate from the 5-hour window and the weekly cap. Agent mode is the usual cause: one prompt fans out into many model requests within seconds.
When you see "Switching to fallback model", the CLI has automatically dropped to a cheaper model to keep going; your results for those turns came from the fallback.
// WHEN IT RESETS
Within a minute. The ceiling is a sliding one-minute window, so simply waiting or slowing the task down clears it. It does not have a reset time in the sense the quota does.
Nothing to wait for on Google's side — this is not an outage and not a fleet-wide condition.
// WHAT TO DO NOW
- 01 Wait 60 seconds and retry. If it recurs, split the task so fewer tools fire per turn.
- 02 Avoid running several Gemini CLI sessions in parallel on the same account — they share the per-minute ceiling.
- 03 Paid Google AI Pro / Ultra tiers have higher per-minute ceilings; check the Quotas page in Google Cloud console for your account's figure.
- 04 If you were silently switched to a fallback model, re-run the important turns once the ceiling clears.
// DOES IT AFFECT MY WEEKLY ALLOWANCE?
No. Rejected requests are not counted, so a burst limit costs you nothing from the 5-hour window or the weekly cap. If the message is "Quota exceeded" instead, that is the allowance — see that page.
// ALSO APPEARS AS
- “Rate limiting detected. Switching to fallback model”
- “429 Too Many Requests”
- “Attempt 1 failed with status 429”