Gemini CLI Limits Explained: Quotas, Reset Time, and What to Do Next

You are midway through a refactor when Gemini CLI reports that a limit has been reached. The right response depends mainly on how you authenticated: a Google account, a personal paid plan, a Workspace license, a Gemini API key, or Vertex AI. This guide helps you identify the applicable daily ceiling, inspect usage with /stats model, and choose whether to wait, upgrade, move to pay-as-you-go, or continue the task in another coding agent.

Gemini CLI daily limits at a glance

The official Gemini CLI quotas and pricing page lists the following maximum model requests per user per day.

Authentication method	Tier or subscription	Maximum model requests per user per day	Important condition
Google account	Gemini Code Assist for individuals	1,000	Gemini CLI chooses across the Gemini model family
Google account	Google AI Pro	1,500	Personal account; fixed-price subscription
Google account	Google AI Ultra	2,000	Personal account; fixed-price subscription
Gemini API key	Free tier, unpaid	250	Flash models only
Gemini API key	Pay-as-you-go	Varies	Depends on pricing tier, model, and token usage
Vertex AI	Express Mode	Varies	Account-specific; free mode lasts 90 days before billing is required
Vertex AI	Regular pay-as-you-go	Varies	Dynamic shared quota or provisioned throughput
Google Workspace	Code Assist Standard	1,500	Requires an assigned organizational license
Google Workspace	Code Assist Enterprise	2,000	Requires an assigned organizational license
Google Workspace	Workspace AI Ultra	2,000	Applies only to the supported Workspace tier

These figures are stated as model requests, not as a guaranteed number of prompts, completed edits, or finished coding tasks. The same official page also says requests are limited per user per minute and remain subject to service availability during periods of high demand. A daily allowance therefore is not the only possible constraint.

First identify which authentication path you are actually using

Your quota is determined by the active authentication and subscription, not simply by the fact that Gemini CLI is installed.

You signed in with a personal Google account

The baseline Gemini Code Assist for individuals allowance is 1,000 model requests per day. Google AI Pro raises the listed maximum to 1,500, and Google AI Ultra raises it to 2,000.

To distinguish a personal account from a Workspace account, the official page suggests opening Google One. A personal account shows the personal dashboard; a Workspace account shows a message saying that you are currently signed in to a Google Workspace account.

You signed in with a Workspace account

The relevant limits depend on the license assigned by your organization: 1,500 requests per day for Code Assist Standard, 2,000 for Code Assist Enterprise, and 2,000 for the supported Workspace AI Ultra tier.

Do not assume that every paid Workspace AI plan increases Gemini CLI quota. The official page says tiers not listed there, including Workspace AI Standard/Plus and AI Expanded, are not supported for this purpose. It also explains that some Workspace Gemini plans apply to web products rather than to the API usage that powers Gemini CLI.

You use a Gemini API key

An unpaid Gemini API key has a lower listed allowance than Google-account login: 250 model requests per day, and requests are limited to Flash models. This is not a quota upgrade from the 1,000-request individual Google-login tier.

A paid API key uses pay-as-you-go pricing. Its quota varies by pricing tier, while cost varies by model and token usage. This path offers more flexibility for continuous work, but it replaces a fixed daily subscription allowance with metered billing.

You use Vertex AI

Vertex AI Express Mode has account-specific quotas and models. The official page says the no-billing period lasts 90 days before billing must be enabled.

Regular Vertex AI is pay-as-you-go. Its quota is governed by dynamic shared quota or pre-purchased provisioned throughput, and cost is based on model and token usage.

Check Gemini CLI usage with /stats model

Run this command inside Gemini CLI:

/stats model

The official documentation says it provides a snapshot of the current session’s token usage and information about limits associated with the active quota. Gemini CLI also presents a model-usage summary when you exit a session.

Use the output for three checks:

1. Confirm which model and quota context the current session is using.
2. Review session token usage before starting another large task.
3. Verify that a newly selected authentication method is active after you sign in again.

Do not treat /stats model as more precise than its actual output. Gemini CLI’s official quota documentation does not promise an exact countdown of how many completed coding tasks remain, and it does not provide a universal quota-reset timestamp. Read the displayed fields literally rather than inferring a number that is not shown.

What time does the Gemini CLI daily quota reset?

The official Gemini CLI quota documentation does not state a fixed reset hour or timezone. It defines the ceilings as requests “per user per day,” but it does not say that every account resets at local midnight, at midnight Pacific Time, or exactly 24 hours after the first request.

That distinction matters because many unofficial answers present one of those possibilities as a guaranteed rule. Based on the official Gemini CLI quota documentation, the defensible answer is that the exact reset clock is undocumented on that page.

If your work can pause, save the current state, retry after the next quota window, and run /stats model again. If your work cannot pause, do not build a schedule around an unverified reset hour; choose a higher supported fixed tier, a paid API key, Vertex AI, or another coding agent instead.

“Limit reached”: choose the next step by situation

There is no single best response. Choose according to urgency, budget model, and whether changing authentication is acceptable.

Your situation	Best first move	Why
The task can wait	Save work and resume after the daily allowance renews	No added cost or account change
You use a personal Google account and repeatedly hit 1,000	Compare Google AI Pro at 1,500 with AI Ultra at 2,000	Higher fixed daily ceiling with predictable subscription cost
Your company manages access	Ask whether Code Assist Standard, Enterprise, or Workspace AI Ultra is assigned	Organizational quota follows the licensed tier
You need long-running work to continue today	Switch to a paid Gemini API key or regular Vertex AI	Pay-as-you-go is the official flexible path after fixed daily limits
You want to avoid metered billing	Pause, reduce the scope, or move the task to another coding agent	Preserves the billing model, but requires context transfer
You use a free API key and need more capacity	Do not merely recreate another free-key workflow; evaluate Google login or paid usage	The free key is listed at only 250 requests per day and Flash only

Waiting is sensible when continuity is not important

Commit or otherwise save the repository state, record the next concrete step, and end the session cleanly. When you return, use /stats model before sending a large prompt so you can see the current session and quota information.

Waiting is a poor choice when a deployment, incident, or long autonomous run must continue. Because the official source does not publish the reset hour, “wait until midnight” is not a reliable operational plan.

A fixed-price upgrade is best for predictable personal or team usage

For a personal Google account, AI Pro and AI Ultra raise the listed daily ceiling without moving to per-token billing. For organizations, Code Assist Standard and Enterprise provide licensed daily limits.

Check that the exact tier is supported before paying. The official page explicitly says Google AI Plus is not among the supported personal tiers, and several Workspace AI plans are also excluded.

Pay-as-you-go is best when stopping costs more than variable billing

A paid Gemini API key is the most direct metered option, while Vertex AI adds Google Cloud controls and enterprise-oriented governance. In both cases, quota varies rather than becoming a single universal daily number.

Metered access still requires cost discipline. The official page notes that many small calls with few tokens can be expensive, so use precise prompts, avoid unnecessary back-and-forth, and monitor session usage with /stats model.

Another coding agent is a workflow fallback, not a quota increase

Moving the task elsewhere does not change the Gemini CLI limit. It simply lets you continue without changing the Gemini account or enabling metered billing.

Before switching, preserve the current branch, uncommitted diff, failing command, acceptance criteria, and the next intended action. This reduces the chance that the new agent repeats work or acts on stale context.

If the error appears before the daily number seems exhausted

A daily limit is not the only explanation. The official Gemini CLI quota documentation also mentions per-user, per-minute limits and service availability during high demand, but it does not publish one universal per-minute number.

Use a low-cost troubleshooting order:

1. Run /stats model and inspect the active model, session usage, and quota information.
2. Check whether you sent a burst of requests in a short period; if so, pause and retry rather than changing plans immediately.
3. Confirm the active account and authentication method. A free API key, personal Google login, and Workspace login have different ceilings.
4. If the daily allowance is genuinely exhausted, choose waiting, a supported fixed tier, or pay-as-you-go.
5. If the message persists despite available quota, treat service availability as a separate possibility and retry later instead of assuming that an upgrade must fix it.

This order avoids paying for the wrong problem. An account upgrade changes the daily ceiling, but it does not guarantee that every transient or per-minute restriction disappears.

Common mistakes that lead to the wrong quota decision

Mistake 1: assuming every paid Gemini plan applies to Gemini CLI

Only the supported tiers listed in the official CLI quota page should be used for planning. A subscription for a Gemini web product does not automatically raise the API quota behind the CLI.

Mistake 2: switching from Google login to a free API key for “more quota”

The published numbers point in the opposite direction: 1,000 requests per day for individual Google login versus 250 for an unpaid API key, with the free key restricted to Flash models.

Mistake 3: treating “per day” as a documented reset time

“Per day” describes the allowance period, not the clock. The official Gemini CLI quota documentation does not identify the timezone or exact hour of renewal.

Mistake 4: interpreting quota as completed-task capacity

The official table measures model requests. A repository-wide migration, a short explanation, and an iterative debugging session are not equivalent units of work, so plan long tasks with margin rather than dividing the daily number by an assumed prompts-per-task figure.

A practical checklist before a long Gemini CLI session

1. Identify whether the session uses personal Google login, Workspace, API key, or Vertex AI.
2. Confirm the exact supported tier and its daily ceiling from the official quota page.
3. Run /stats model and record the current session information.
4. Save or commit a clean checkpoint before a long autonomous task.
5. Decide in advance what happens at the limit: wait, upgrade, switch to paid usage, or transfer to another agent.
6. If using pay-as-you-go, set a cost-monitoring routine and favor precise prompts over repeated small calls.
7. Do not depend on a specific reset hour unless Google documents it for your account and authentication path.

With those decisions made before the run, a limit message becomes a routing choice rather than an emergency.

Official references

• Gemini CLI: Quotas and pricing
• Gemini CLI command reference for /stats