Gemini usage limits, explained

The short answer

Which Gemini limit are you hitting?

“Gemini limit” can mean four or five different meters. They don’t share a pool, they don’t reset together, and checking one tells you nothing about the others. So before anything else, figure out where the “you’ve reached your limit” message came from.

If you code with Gemini, the row that matters most is probably not the first one. The Gemini app’s allowance has nothing to do with your CLI quota, even when both sit on the same Google account. A sloth with two bowls of leaves can empty one and still have the other.

Gemini app limits by plan

Google’s help article “Gemini Apps limits & feature availability” is the official source. It covers personal Google accounts; work and school accounts have a separate article. Here’s what it says:

Notice what’s missing: a number. Google doesn’t say how many prompts “standard” is, and it doesn’t give per-model daily counts for Deep Research, image, video or music generation on this page. If a blog post tells you “Pro gets exactly N prompts a day”, it isn’t quoting Google’s current help center. The page also says limits “may change without notice, including due to capacity constraints.”

What uses more of your allowance

Google lists the things that cost more: longer chats, analyzing files, Deep Research, and generating media, with video “generally” using the most. Its practical tip is to start a new chat for each new topic, so Gemini isn’t rereading a long history every turn. For people pasting big code files into the app, that tip does a lot of work.

The context window is a different limit

The context window is how much one conversation can hold, not how much you can use in a day. A free account’s 32k tokens fills up fast with source files; Google notes that going past the window can make Gemini miss details. Running out of context and running out of usage feel similar in the moment, but the fixes are different: a fresh chat for the first, waiting or upgrading for the second.

When the Gemini app says you’re out

Per Google’s help center:

• You get a warning first. The app notifies you when you’re close to your limit, based on your plan.
• It tells you when it comes back. Once you hit the limit, the app says when your limit refreshes.
• Subscribers can keep going on Flash-Lite. If you have a Google AI subscription and reach your limit, you can continue the conversation with Flash-Lite.
• Otherwise: wait or upgrade. Google’s two options for a five-hour or weekly limit are upgrading your plan or waiting for the refresh. Google’s plans page also mentions AI credits for extra usage; availability varies by country.

Gemini CLI: the June 2026 change

This is the part most older articles get wrong. On May 19, 2026, Google’s Developers Blog announced that Gemini CLI is transitioning to Antigravity CLI. The key date was June 18, 2026:

• Free individual tier, Google AI Pro and Ultra: Gemini CLI and the Gemini Code Assist IDE extensions stopped serving requests for these users. Google’s own phrasing: they “will stop serving requests for Google AI Pro and Ultra.” The no-cost Gemini Code Assist for individuals is affected too.
• Gemini Code Assist Standard and Enterprise: “your access remains unchanged.” Organizations keep Gemini CLI and Code Assist with the latest models and updates.
• API keys: “Gemini CLI will remain accessible via paid Gemini and Gemini Enterprise Agent Platform API keys.”
• Gemini Code Assist for GitHub: no new installs for consumer tiers from June 18, with requests stopping in the following weeks. Organizations on Google Cloud Standard or Enterprise licenses aren’t affected.

The Gemini CLI docs and GitHub README now open with the same notice: for unpaid tier and Google One users, “Gemini CLI was replaced by Antigravity CLI on June 18th, 2026.” They also say Gemini CLI “has not been optimized for Gemini 4 Argon” and point you to Antigravity for that model.

So if you signed in to Gemini CLI with your personal Google account and it suddenly stopped working this summer, that’s not a quota. It’s the transition. More waiting won’t bring it back.

Gemini CLI quotas by sign-in

Google’s “Quota and pricing” page for Gemini CLI (last updated June 18, 2026) lists daily request limits per user. It still shows the consumer tiers, so we include them, with a note on what still applies:

A few details from the same page that are easy to miss:

• There’s a per-minute limit too. “Requests are limited per user per minute and are subject to the availability of the service in times of high demand.” Google doesn’t publish that number, so if you’ve seen “60 requests per minute” quoted, it isn’t on the current page.
• Requests aren’t prompts. “Model requests will be made across the Gemini model family as determined by Gemini CLI.” One prompt that reads files and runs tools can make several model requests, so a daily limit of 1,500 isn’t 1,500 questions.
• Some plans don’t cover the CLI at all. Google AI Plus, Workspace AI Standard/Plus and Workspace AI Expanded aren’t supported, and Gemini for Workspace plans don’t cover the API usage that powers the CLI.

Gemini Code Assist quotas

For organizations on Standard or Enterprise, Google Cloud’s Gemini quotas page (last updated October 13, 2026) is the reference:

Two things worth knowing. The agent-mode and CLI limit is combined across those two interfaces and across all model versions and families, so heavy agent mode in your IDE eats into your terminal budget. And the per-minute limit for agent mode and the CLI exists but, again, isn’t published as a number. Gemini Code Assist on GitHub has its own quota of at least 100 pull request reviews per day per installation, which doesn’t count toward the others.

Gemini API rate limits

If you run Gemini CLI or your own tools with an API key, you’re on the Gemini API’s rate limits. Google’s rate-limits page (last updated October 15, 2026) explains how they work:

• Three main dimensions: requests per minute (RPM), input tokens per minute (TPM) and requests per day (RPD). Some models add images per minute or tokens per day.
• Per project, not per key. “Rate limits are applied per project, not per API key.” Making a second key in the same project doesn’t double anything.
• Daily reset at midnight Pacific. RPD quotas reset at midnight Pacific time, which is morning in Europe and afternoon in much of Asia.
• Tiers by spend. Free; Tier 1 once you link a billing account; Tier 2 after $100 in paid spend and 3 days since your first successful payment; Tier 3 after $1,000 and 30 days. Each tier has a billing cap.
• No free-tier numbers on the page. Exact limits depend on the model and your tier, and Google shows your current ones in AI Studio at aistudio.google.com/rate-limit.

Going over any limit triggers a rate-limit error. For spend-based limits the page names the error as 429 RESOURCE_EXHAUSTED.

Antigravity, briefly

Since personal-account CLI users were moved to Antigravity CLI, its limits are now the ones many solo developers live with. Google’s Antigravity plans page describes them without numbers:

• Google AI Ultra: “The highest, most generous quota, refreshed every five hours,” plus the highest weekly rate limits.
• Google AI Pro: “A high, generous quota, refreshed every five hours until the weekly limit is reached,” plus a higher weekly rate limit.
• Everyone else: “A meaningful quota, refreshed weekly.”

Pro and Ultra subscribers can buy AI credits for usage beyond the baseline and choose whether Antigravity uses them automatically. Baseline quota usage is shown on Antigravity’s Settings page. The page notes that limits depend on capacity and are “subject to modification.” Treat any precise Antigravity number you see elsewhere as unofficial.

How each limit resets

The five-hour-plus-weekly shape will look familiar if you also use Claude Code or Codex. See when Claude Code resets and Codex usage limits for how their windows work, or the glossary if “rolling window” and “weekly cap” blur together.

How to check your Gemini usage

Gemini app

Google’s help center: go to gemini.google.com, select Settings at the bottom left, then Usage limits. The Gemini Mac app has the same pane under Settings → Usage limits.

Gemini CLI

/stats model

Google’s commands reference says /stats model shows “model-specific usage statistics, including token counts and quota information.” Plain /stats (or /stats session) shows session duration, tool calls and performance; /stats tools breaks usage down by tool. A usage summary also prints when you exit a session. If you’re signed in with Google, /upgrade opens the Code Assist upgrade page for higher limits.

Gemini API

Your current rate limits are in Google AI Studio at aistudio.google.com/rate-limit. Vertex AI quotas live in the Google Cloud console.

Antigravity

The Settings page shows baseline quota usage across models.

What to do when you hit a limit

1. Check which meter it is. An app limit, a CLI quota and a 429 from the API are three different problems.
2. In the Gemini app: switch to Flash-Lite if you’re a subscriber, start fresh chats to use less, or wait for the five-hour refresh. Upgrade if the weekly limit keeps stopping you.
3. In Gemini CLI on Standard or Enterprise: run /stats model, then wait for the daily quota or talk to your admin. Google’s quota page calls pay-as-you-go (a Gemini API key or Vertex AI) “the recommended path for uninterrupted access.”
4. In Gemini CLI on a personal account: it isn’t a quota. Move to Antigravity CLI, or use a paid API key.
5. With an API key: slow down for per-minute limits, wait until midnight Pacific for daily ones, or move up a tier by linking billing.
6. Or switch tools for a while. If you also have Claude Code or Codex, now is a good time. See Claude Code vs Codex limits.

Why usage drains faster than you expect

In the app, Google’s list is short: long chats, file analysis, Deep Research and media generation. Every turn of a long chat carries the whole conversation, so the hundredth message costs more than the first. In the CLI, it’s requests, and an agent that reads a dozen files, runs a test and edits three more has made many requests from one prompt. Neither is a bug. Both reward small, focused sessions, which is also how sloths prefer to work.

How Quotchi reads Gemini

Honestly, and with limits of our own. Quotchi is built for Claude Code and Codex. Gemini is optional, and here’s exactly what that means:

• Gemini Mac app only. Quotchi reads the usage bars in the Gemini desktop app’s Usage limits pane. It doesn’t read the Gemini CLI, Antigravity, Code Assist or your API quota.
• Off by default. You turn on Read the Usage limits pane in Quotchi → Settings → AI tools → Gemini, and grant macOS Accessibility permission.
• The pane has to be open. Leave Gemini’s English Usage limits pane open. If the numbers look old, refresh inside Gemini; Quotchi reads what the pane shows.
• App usage, nothing else. No reset times and no token stats, because the pane doesn’t show them. The reader is experimental and tested on specific Gemini app versions, listed on the compatibility page.
• No chats. The reader looks only at the usage bars and never reads chat windows. See the privacy guide.

Once it’s on, Gemini sits next to Claude and Codex in the panel, and quotchi wait gemini works in scripts like the other tools. Setup steps are in the help center.

Common myths

“Gemini CLI is free with any Google account.”

It was. Since June 18, 2026, free individual accounts, AI Pro and Ultra are served by Antigravity CLI instead.

“My Google AI Pro plan covers the app and the CLI from one pool.”

The app allowance and any coding quota are separate meters with separate resets.

“1,500 requests means 1,500 prompts.”

The CLI can make several model requests per prompt.

“A new API key gets me fresh limits.”

API rate limits are per project, not per key.

“Google publishes how many prompts Pro gets.”

Not on its current help page. It publishes multipliers and says limits can change.

What changed recently

Sources

Official pages we read on October 17, 2026. If anything here disagrees with them, they win.

• Gemini Apps Help: Gemini Apps limits & feature availability
• Google Developers Blog: Transitioning Gemini CLI to Antigravity CLI (May 19, 2026)
• Gemini CLI docs: Quota and pricing
• Gemini CLI docs: CLI commands
• Gemini CLI on GitHub
• Google Cloud: Gemini quotas and limits
• Gemini API docs: Rate limits
• Google Antigravity docs: Plans
• Google AI plans

FAQ