Requests per day
The daily pool. It refills at midnight Pacific Time (00:00 America/Los_Angeles) — 08:00 UTC in winter (PST, UTC−8), 07:00 UTC in summer (PDT, UTC−7). This is the reset the countdown above tracks.
Google Gemini · API quota resets
The Gemini API counts requests per day (RPD) against the clock in America/Los_Angeles. When the Pacific day rolls over, your daily quota refills — whether that's noon, 3 PM, or 4 AM where you live. This page tracks that reset live.
Time until next daily reset
Counts down to 12:00 AM Pacific Time, when RPD quotas refill for every Gemini API project.
Measuring the Pacific day…
Gemini API usage is metered in three dimensions. Only one of them runs on a daily clock — and that clock is Pacific Time. All limits apply per Google Cloud project, not per API key.
The daily pool. It refills at midnight Pacific Time (00:00 America/Los_Angeles) — 08:00 UTC in winter (PST, UTC−8), 07:00 UTC in summer (PDT, UTC−7). This is the reset the countdown above tracks.
A short rolling window, not a calendar reset. Exceed it and you get a 429 — but the capacity comes back within about a minute, not at midnight. Bursts trip this; steady daily volume doesn't.
Input tokens per minute (some models also cap tokens or images per day or minute). Like RPM, it recovers on a per-minute basis — a few very large prompts can exhaust it even at low request counts.
Google sets limits per model and per usage tier, and no longer publishes one fixed table of numbers — your project's live limits are shown in AI Studio. What stays constant is the shape: the midnight-Pacific reset matters most on the free tier, where the daily cap is the binding constraint.
| Tier | How you get it | Daily (RPD) behavior | At the reset |
|---|---|---|---|
| FreeNo billing attached | Create an API key in Google AI Studio in an eligible region — no card required. | Hard daily capSmall per-model RPD pool; the daily limit is usually what stops you first. Free-tier content may be used to improve Google products. | Full RPD pool refills at midnight Pacific. Nothing carries over, nothing rolls forward. |
| Tier 1Paid | Link a Google Cloud billing account to the project in AI Studio. | Much higher limitsRPM and TPM jump dramatically and daily caps are far larger or absent for most models; per-minute limits and spend caps become the practical ceiling. | Same midnight-Pacific clock for any daily dimension that still applies — but most workloads stop noticing it. |
| Tier 2 & 3Paid, high volume | Unlocked automatically as cumulative paid spend and account age grow — Google publishes the current thresholds in its billing documentation. | Highest limitsDesigned for production traffic; daily caps are effectively replaced by per-minute limits and billing-account spend caps. | The Pacific-midnight reset rarely binds; watch RPM/TPM and your spend caps instead. |
Exact numbers change as Google adjusts models and tiers, and preview/experimental models run tighter limits than stable ones. Treat any fixed RPM/TPM/RPD table copied from a blog as a snapshot — the authoritative figure for your project is the one AI Studio shows today.
Your limits live with your project, not with this page. These are the official places Google exposes them:
Your project's actual RPM, TPM, and RPD for every model, plus how much of today's daily quota you've used. This is the number that matters.
aistudio.google.com/rate-limitGoogle's own explanation of the limit dimensions, per-project scoping, tier structure, and the midnight-Pacific daily reset rule.
ai.google.dev/gemini-api/docs/rate-limitsHow linking billing moves a project to Tier 1, and the spend and account-age thresholds that unlock Tiers 2 and 3 with higher limits.
ai.google.dev/gemini-api/docs/billingWhat each model costs on paid tiers, and which models are available on the free tier at all before you plan around their daily caps.
ai.google.dev/gemini-api/docs/pricingThe questions people actually search for when a 429 shows up at the worst moment.
At midnight Pacific Time — 00:00 in America/Los_Angeles. That's 08:00 UTC while Pacific Standard Time (UTC−8) is in effect and 07:00 UTC during Pacific Daylight Time (UTC−7). The countdown at the top of this page targets that exact moment.
No. The daily reset is anchored to Pacific Time for every project, everywhere. In London the refill lands at 08:00 in winter, 07:00 in summer; in Shanghai at 16:00 during PST months, 15:00 during PDT months. For your exact zone, use the “Reset moment (your time)” readout in the countdown panel above — it converts the next Pacific midnight for you.
Requests start failing with HTTP 429 (RESOURCE_EXHAUSTED) and keep failing until the midnight-Pacific reset. No amount of retry backoff helps against an exhausted daily pool — your options are to wait for the reset, move traffic to a different project, or upgrade the project to a paid tier.
No. Requests-per-minute and tokens-per-minute are short rolling windows, not daily counters. If a 429 clears after waiting a minute, you hit RPM or TPM. If it persists for hours, you hit RPD — and only the midnight-Pacific reset (or a tier upgrade) clears it.
Per project. Every API key inside one Google Cloud project draws from the same RPM, TPM, and RPD pools. Issuing a fresh key in the same project changes nothing; a genuinely separate project gets its own quota (and its own billing and tier status).
Daylight saving time. The reset is always midnight Pacific local time, but Pacific alternates between UTC−8 (PST) and UTC−7 (PDT). When the US springs forward or falls back, the reset shifts by one hour relative to UTC and to any timezone that doesn't change on the same date.
Mostly. Linking billing (Tier 1) raises RPM/TPM dramatically and makes daily caps far larger or removes them for most models, so the midnight boundary stops being the constraint; higher tiers go further. Any daily dimension that does still apply keeps the same midnight-Pacific reset — upgrading changes the size of the pool, not the clock.
Different system. This page covers the Gemini API (AI Studio keys and Google Cloud projects). The consumer Gemini app and its Google AI plan limits are metered separately, and Google doesn't run them on the API's published midnight-Pacific daily clock. Don't plan app usage around the API reset — or API usage around the app's limits.
On the AI Studio rate-limit page, which shows your project's live per-model limits and current usage. Google no longer publishes a universal per-model table, so the figures shown for your project are the authoritative ones.