Gemini CLI free tier limits, explained

As of 22 October 2026, Gemini CLI has no free tier for personal Google accounts. On 18 June 2026 Google stopped serving it for the free Gemini Code Assist for individuals (1,000 requests a day) and for Google AI Pro and Ultra. It still runs on Code Assist Standard (1,500 requests a day), Enterprise (2,000) and paid Gemini API keys.

Pages written before June still quote the old 1,000 a day as if it were current.

What happened to the Gemini CLI free tier?#

Google announced the change on 19 May 2026: "On June 18, 2026, Gemini CLI and Gemini Code Assist IDE extensions will stop serving requests for Google AI Pro and Ultra", and for people using it at no charge through Gemini Code Assist for individuals (Google Developers Blog).

The same post says what stays: enterprise access on a Code Assist Standard or Enterprise licence "remains unchanged", and "Gemini CLI will remain accessible via paid Gemini and Gemini Enterprise Agent Platform API keys". For everyone else, Google's answer is Antigravity CLI.

The Gemini CLI repository opens with the same news: "Unpaid tier and Google One users: Gemini CLI was replaced by Antigravity CLI on June 18th, 2026" (GitHub).

Gemini CLI limits by sign-in, as of October 2026#

Google's quota page still lists every tier, including the ones that stopped. Its last update is dated 18 June 2026 (Gemini CLI: Quotas and pricing). Here is that list next to what the announcement says about each:

Every tier also has a per-minute limit. The quota page gives no number for it: "Requests are limited per user per minute and are subject to the availability of the service in times of high demand."

Two more things to know. A request isn't one prompt: one prompt can make several model requests while the agent reads files and runs tools. And the README says Gemini CLI "has not been optimized for Gemini 4 Argon"; to use that model, Google points you to Antigravity.

When do Gemini CLI limits reset?#

For a Gemini API key, Google's rate-limit page is clear: "Requests per day (RPD) quotas reset at midnight Pacific time" (Gemini API: Rate limits). That's 09:00 in Amsterdam for most of the year.

For Code Assist Standard and Enterprise, the quota page gives the daily numbers but no reset time. If you hit the daily cap on a licence, assume you're out until the next day and check with your Google Cloud admin.

What "quota exceeded" means#

It means you've used up a limit for the way you signed in: requests per minute, or requests per day. On an API key, Google's page says going over any limit "will trigger a rate limit error", and names 429 RESOURCE_EXHAUSTED as the error for its spend-based limits.

A common case on Google's developer forum, and an estimate rather than official guidance: a key that reports limit: 0 has no quota for that model at all, usually because the project has no billing account. Google's tier table says what moves you up: Tier 1 needs "an active billing account", Tier 2 "Paid $100 + 3 days from first successful payment", Tier 3 "Paid $1,000 + 30 days".

If you signed in with a personal Google account and Gemini CLI refuses every request, it isn't a quota at all. It's the 18 June change. Signing Gemini CLI in to a Code Assist Standard or Enterprise account instead takes /auth signout and a GOOGLE_CLOUD_PROJECT (the steps).

What to do when you hit the limit#

Free and manual first:

1. Wait for the reset. For an API key, midnight Pacific time.
2. Slow the per-minute rate. Fewer agents at once on the same sign-in, or smaller tasks.
3. Add billing to your API key's project. It moves the key from the free tier to Tier 1 and raises its limits. You then pay per token.
4. Move to Antigravity CLI if you used a personal Google account. Google's migration page says it still reads GEMINI.md and AGENTS.md, converts Gemini extensions with agy plugin import gemini, and moves MCP servers to a new mcp_config.json (Antigravity docs). How the two CLIs differ is in Gemini CLI vs Antigravity CLI, and Gemini CLI alternatives covers agents from other providers.
5. Use another agent for the rest of the day. If you also pay for Claude or ChatGPT, our Claude Code vs Gemini CLI comparison covers how the limits differ.

Use only accounts and keys you own and are entitled to use. Gemini CLI's terms depend on how you sign in, and creating extra accounts to get around a limit is not what any of this is for.

How Shuriken handles more than one Gemini source#

Shuriken runs Gemini CLI and Antigravity CLI among its 67 coding agents. For Gemini CLI, every Google sign-in and every Gemini API key you add is its own source. Each Google sign-in lives in its own folder (Shuriken sets Gemini CLI's GEMINI_CLI_HOME per source), so a second sign-in never replaces the first. Your GEMINI.md, commands, extensions and skills are shared across them.

If you hold more than one Gemini source, say two paid API keys, or a sign-in and a key, add each one. When the source in use reports a limit, Shuriken moves your Gemini chats to another of your Gemini sources at the end of their current turn, on every linked device; that's our source switching. Sign-ins come first. A chat on a sign-in moves onto an API key only if you turn on "Switch to API keys", which is off by default and has a $25 spend cap per fallback run; a chat already running on a key can move to your other key.

What Shuriken can't do is bring back the free tier: a personal Google sign-in won't serve Gemini CLI there either. If you hold more than one paid Gemini source, download Shuriken and add them.