// WHAT IT MEANS
Your allowance for the current window is used up. Before May 2026 Gemini CLI counted prompts per day (1,000 on a personal Google account, 1,500 on AI Pro, 2,000 on Ultra); since then it uses compute-based 5-hour windows with a weekly cap layered on top. "RESOURCE_EXHAUSTED" and "Quota exceeded for quota metric" are the same condition reported by the API layer.
The quota counts model requests, not how many times you hit enter. In agent mode a single prompt can trigger dozens of requests internally, which is why the bar drops faster than you would expect.
// WHEN IT RESETS
5-hour window: 5 hours after the first request that opened it. Weekly cap: Google has not spelled out whether it is a fixed weekly time or a rolling 7 days — go by what your account shows.
Google does not do fleet-wide resets the way OpenAI does; its changes are almost all rule changes, which we log below.
// WHAT TO DO NOW
- 01 Run /stats model to see token use for this session — note it covers only the current session, not your account total.
- 02 Reduce request fan-out: fewer tools per turn, smaller tasks, and avoid loops that re-plan on every step.
- 03 On a personal Google account, remember that since June 2026 free and Google One users are on the Antigravity CLI rather than Gemini CLI; the quota tiers work the same way.
- 04 Account-level quota and paid-tier limits live in the Google Cloud console under Quotas, and on the Google One / AI subscription page.
// DOES IT AFFECT MY WEEKLY ALLOWANCE?
If the 5-hour window is what emptied, the weekly cap still has room. If both are exhausted the message is the same — your account page is the only place that distinguishes them.
// ALSO APPEARS AS
- “Quota exceeded for quota metric”
- “RESOURCE_EXHAUSTED”
- “You have exhausted your daily quota”