Gemini CLI limits: free tier, daily caps, resets

What are Gemini CLI’s limits right now?

As of October 2026, Gemini CLI’s limits depend on how you sign in, and a personal Google account is no longer one of the options. The routes left are a Gemini API key, Vertex AI and a company Gemini Code Assist license. With Code Assist Standard or Enterprise, Google’s quotas page lists 1,500 or 2,000 requests per user per day, though Google stopped selling new ones on October 9, 2026. With a paid API key the limits follow your usage tier. An unpaid key is the open question, because Google’s pages give different numbers for it and say little about whether it still works.

The personal route ended on June 18, 2026. Google says Gemini CLI stopped serving requests for the Gemini Code Assist for individuals (free), Google AI Pro and Google AI Ultra tiers that day. You can no longer use Login with Google with a personal account to reach it, per its deprecation notice.

Some of the project’s own pages have not caught up. The README still advertises a free tier of 60 requests a minute and 1,000 requests a day with a personal Google account. The authentication guide still recommends logging in with your personal Google account, and the Plans page still offers a free plan that authenticates with a Google Account. The quota page still lists rows for Gemini Code Assist for individuals (1,000 requests), Google AI Pro (1,500) and Google AI Ultra (2,000), while a banner on it says Gemini CLI was replaced by Antigravity CLI on June 18, 2026 for the “unpaid tier” and Google One users. Where the pages disagree, follow the dated notice. Our Claude Code vs Codex vs Gemini CLI comparison records the same change.

Does Gemini CLI still have a free tier?

Not through a personal Google account. For an unpaid Gemini API key, Google’s pages are mixed. The quota page lists “Log in with Gemini API Key (unpaid)” at 250 model requests per user per day, Flash models only. The README advertises a free API key at 1,000 requests a day with Gemini 3, a “mix of flash and pro.” Against that, the quota page banner says Gemini CLI was replaced for the “unpaid tier.” The enterprise section of Google’s announcement says Gemini CLI remains accessible through paid API keys, and Google Cloud’s Gemini CLI page says it supports a Gemini API key to pay as you go. None of these pages names unpaid API keys as withdrawn, and the numbers do not agree. Test your key, and check its limits in Google AI Studio, before you plan around it.

If you use a free key, know the free tier’s own limits. Google’s pricing page lists free access model by model. It marks Gemini 4 Argon and Gemini 3.1 Pro Preview as not available on the free tier, and lists models such as Gemini 3.8 Flash as free of charge. Google’s rate limits page does not list per-model request limits. It says limits depend on your usage tier, are shown in Google AI Studio, and are not guaranteed, so check your own there.

The free tier can carry a privacy cost. Google’s Gemini API terms say that on unpaid quota Google uses the content you submit, and the responses, to improve its products, that human reviewers may read it, and that you should not submit sensitive, confidential or personal information. In the European Economic Area, Switzerland and the United Kingdom, the paid-services terms apply to unpaid quota as well. On paid quota, Google says it does not use your prompts or responses to improve its products. Weigh that before you point a free key at a private repository.

How many requests per day does Gemini CLI allow?

Count requests, not prompts. Google’s quotas page says that with Gemini CLI one prompt might result in multiple model requests, and that the daily limit is aggregated across every model you use. By route:

Unpaid API key. The quota page says 250 requests per user per day, Flash models only, and the README says 1,000 a day. See the open question above.

Paid API key. You pay per token, and your limits depend on your usage tier. Tier 1 needs an active billing account, Tier 2 needs $100 paid plus 3 days. Tier 3 needs a total spend of $1,000 plus 30 days, according to the rate limits page.

Vertex AI. Express mode needs no billing for 90 days, with limits that vary by account. Regular mode uses a dynamic shared quota or provisioned throughput, per the quota page.

Gemini Code Assist Standard or Enterprise. 1,500 or 2,000 requests per user per day, per the quotas page. Google stopped selling new Standard and Enterprise subscriptions on October 9, 2026, and existing ones keep working until the end of their contract, per its notice.

When do Gemini CLI limits reset?

For a Gemini API key, Google says requests-per-day quotas reset at midnight Pacific time, and that limits apply per project, not per API key, so a second key in the same project adds nothing. Requests per minute and tokens per minute are limits too, and exceeding any one of them triggers a rate limit error. See how rate limits work.

For Code Assist licenses, Google says only that once the daily maximum is reached no more requests can be made “until the quota resets,” and gives no time on that page. Gemini CLI itself tells you. When a model’s quota runs out it prints “Usage limit reached for” that model and, when it knows, “Access resets at” a time. It then points to /stats model for details and /model to switch models, per the open-source CLI source. /stats model shows token counts and quota information, according to the command reference.

What do you do when Gemini CLI says the limit is reached?

First work out which problem you have. “Usage limit reached” is your quota. “We are currently experiencing high demand” is capacity on Google’s side, and Gemini CLI then offers to keep trying or to switch models, per the CLI source. A 429 “Resource exhausted” API error means you exceeded a request limit, and the Gemini CLI FAQ says to check your usage in Google AI Studio or your Cloud dashboard, batch prompts or add delays, or request a quota increase. Then pick a path:

Wait. The CLI shows the reset time when it knows it.

Switch model. /model changes the model, and the dialog Gemini CLI shows when a model runs out offers a switch to a fallback model, per the CLI source.

Go pay-as-you-go. The quota page recommends a Gemini API key or Vertex AI when you keep hitting daily limits. It adds that paying per token can cost more for many small calls, and is the only way to make sure a quota does not interrupt your work. See pay as you go.

Use another agent for the rest of the day. The work does not belong to one CLI, and Codex and Claude Code have their own allowances.

Can Google AI Pro or Ultra raise Gemini CLI limits?

Not any more. Google AI Pro and Ultra stopped being served in Gemini CLI on June 18, 2026, per the deprecation notice, even though the Gemini CLI FAQ still says subscribers automatically get higher limits. Google points consumer accounts to Antigravity CLI. Its pricing page lists a $0 individual plan with “basic weekly rate limits,” and Google AI Pro and Ultra plans with higher rate limits. Those are Antigravity’s limits, not Gemini CLI’s daily request caps, so read that page before you compare.

Run Gemini CLI next to your other agents

In SwarmPane, Gemini CLI runs with a Gemini API key, so the API key limits above are the ones you get. Agent Grid runs up to 16 agents in real terminals, with Gemini CLI and Antigravity CLI beside Claude Code, Codex and the others, each pane showing whether it is working, waiting for you or done. Broadcast sends one prompt to several panes and says how many took it. When one CLI is out of requests, the others keep working.

SwarmPane runs the agent CLIs and accounts you already have, so Google bills its own usage separately. Start with a 7-day trial for $1 and cancel before it ends if it is not for you.