Home / Blog / Pricing

Pricing

Gemini CLI "429 RESOURCE_EXHAUSTED": What It Means and How to Fix It

Gemini CLI throwing 429 RESOURCE_EXHAUSTED? Per-minute vs daily quota, why it falls back to Flash, and the fixes: wait, switch auth, API key, or paid plan.

By Engineer Foodie Desk ✓ Sources checked Oct 30, 2026

9 min read

Quota table checked Oct 30, 2026 on Gemini CLI’s quota and pricing page; rate-limit rules Oct 30, 2026 on Gemini API rate limits and Google Cloud’s error code 429 page; error text from Gemini CLI’s official GitHub repo (verified Oct 29, 2026). Other dates under Sources. No affiliate links on this page (how we work).

The exact error: "code": 429 … "status": "RESOURCE_EXHAUSTED"

This is the message Gemini CLI users pasted into the project’s GitHub issues:

[API Error: [{
    "error": {
      "code": 429,
      "message": "Resource exhausted. Please try again later. Please refer to https://cloud.google.com/vertex-ai/generative-ai/docs/error-code-429 for more details.",
      "errors": [
        {
          "message": "Resource exhausted. Please try again later. Please refer to https://cloud.google.com/vertex-ai/generative-ai/docs/error-code-429 for more details.",
          "domain": "global",
          "reason": "rateLimitExceeded"
        }
      ],
      "status": "RESOURCE_EXHAUSTED"
    }
  }
  ]]
  Possible quota limitations in place or slow response times detected. Switching to the gemini-2.5-flash model for the rest of this session.

Source: gemini-cli issue #10722 and issue #10889 (verified Oct 29, 2026). Those are 2025 builds (CLI 0.8.1 and 0.8.2), so the model name in the last line may differ on your version.

What it means, as of Oct 30, 2026: you’ve hit a rate limit. Google’s Cloud docs say a 429 comes back when “the number of your requests exceeds the capacity allocated to process requests” (verified Oct 30, 2026). Either you used up a per-minute allowance (wait a minute), a daily one (wait for the reset), or a spend cap (wait about ten minutes or raise your tier). Which one depends on how you sign in. And if you used to sign in with a personal Google account, read the next section first, because that door closed in June.

We haven’t reproduced this error. Everything below is from Google’s docs and the Gemini CLI repo, quoted and dated.

First check: are you still on “Login with Google”?

Issue #10722 came from a Google account sign-in, and #10889 from someone who had just upgraded to Pro. That route is gone for individuals. Google’s Code Assist page says: “Starting June 18, 2026, Gemini Code Assist IDE extensions stopped serving requests for the Gemini Code Assist for individuals, Google AI Pro, and Google AI Ultra tiers. This also applies to usage of Gemini CLI. As part of the deprecation, you can no longer use the Login with Google option to access the IDE extensions or Gemini CLI” (verified Oct 29, 2026).

The Gemini CLI quota page opens with the same news: “Unpaid tier and Google One users: Gemini CLI was replaced by Antigravity CLI on June 18th, 2026” (verified Oct 30, 2026). Then, a few lines down, the same page still lists Google AI Pro at 1,500 and Ultra at 2,000 requests a day. That’s stale, and it’s maddening: it’s the page you land on mid-error, and it gives two answers. Go with the banner. More on what changed: Gemini CLI free tier limits.

So in 2026 a 429 in Gemini CLI almost always means one of three setups: a Gemini API key, Vertex AI, or a company Gemini Code Assist license.

Per-minute vs daily quota

The CLI’s quota page sets daily caps and adds: “Requests are limited per user per minute and are subject to the availability of the service in times of high demand” (verified Oct 30, 2026). So there are two walls, plus a third for paid API keys.

Per-minute. The Gemini API measures “Requests per minute (RPM)”, “Tokens per minute (input) (TPM)” and “Requests per day (RPD)”, and “exceeding any of them will trigger a rate limit error” (verified Oct 30, 2026). If the error clears after a minute or so, this was it. Google doesn’t publish per-minute numbers on these pages; your project’s live limits “can be viewed in Google AI Studio”.

Daily. The documented daily caps by sign-in (quota page, verified Oct 30, 2026):

  • Gemini API key, free tier: “250 maximum model requests / user / day”, “Model requests to Flash model only”.
  • Gemini API key, pay-as-you-go: “Varies”.
  • Vertex AI Express mode: “Quotas and models are specific to your account and their limits vary”, with “90 days before you need to enable billing”.
  • Code Assist Standard: 1,500 a day; Enterprise: 2,000 a day.

Spend-based (paid API keys). The Gemini API also caps spend over “a rolling 10-minute window”: $10 on Tier 1, $50 on Tier 2, $200 on Tier 3, none on Free. “If you hit a spend-based rate limit, the API returns a 429 RESOURCE_EXHAUSTED error” (verified Oct 30, 2026).

The “subject to the availability of the service in times of high demand” clause is the George Costanza “I was in the pool!” of quota docs: sometimes it’s smaller than you expected, and the reason is the conditions, not you. Google publishes nothing more specific.

The Flash fallback line: what it does

That last line, “Possible quota limitations in place or slow response times detected. Switching to the gemini-2.5-flash model for the rest of this session.”, is Gemini CLI’s fallback. Older versions showed it when the Pro model returned 429, and moved the session to a Flash model.

It didn’t always work. In #10722 the reporter says the switch was “inconsistent and sometimes the tool gets stuck in an error loop” (CLI 0.8.1, Google account sign-in). In #10889 a user who had just upgraded to Pro says the CLI “does not do what it says, i.e. change to the flash model, and keeps erroring”, and adds: “Even when I run the CLI with ‘gemini -m gemini-2.5-flash’ I still get the error” (CLI 0.8.2). Both issues are closed (verified Oct 29, 2026).

A message announcing a switch that doesn’t happen is Michael Scott walking out of the office yelling “I DECLARE BANKRUPTCY!” Saying it out loud isn’t the same as doing it.

API-key users had a different problem: no fallback at all. A 2026 pull request, #19786, describes API key users hitting RESOURCE_EXHAUSTED (429) on Pro models who “get no fallback dialog — only a raw error”, because the fallback was limited to Google sign-in. That PR was closed in March 2026 without being merged (verified Oct 30, 2026). So don’t count on an automatic switch with an API key. Switch the model yourself.

Fixes, in order

Legit fixes only, from Google’s own docs. Nothing here involves extra accounts or tricks; and spinning up new API keys won’t help anyway, because Gemini API “Rate limits are applied per project, not per API key” (verified Oct 30, 2026).

  1. See which limit you hit. Run /stats model. The quota page says it shows “your current session’s token usage, as well as information about the limits associated with your current quota” (verified Oct 30, 2026). On an API key, check your project’s limits in Google AI Studio.
  2. Wait and retry. For a per-minute or spend limit, Google’s advice is “Wait and retry after a short period” (Gemini API) and “a retry strategy by using truncated exponential backoff” (Cloud docs). For the daily limit on an API key, “Requests per day (RPD) quotas reset at midnight Pacific time” (verified Oct 30, 2026). For Vertex and Code Assist, the CLI page publishes no reset time.
  3. Use a lighter model or a lighter prompt. “Rate limits are more restricted for experimental and preview models,” and Google suggests “smaller context windows or shorter outputs” for spend limits (verified Oct 30, 2026). On a free API key, you’re on Flash anyway.
  4. Change how you sign in. If you were on Login with Google with a personal or Google AI Pro account, that’s over; use a Gemini API key, Vertex AI, or move to Antigravity CLI, which Google names as the replacement. The CLI docs call pay-as-you-go via “a Gemini API key or Vertex AI” “the recommended path for uninterrupted access” (verified Oct 30, 2026).
  5. Go paid. On an API key, linking a billing account moves the project to Tier 1, and “Tier upgrades from the Free to Tier 1 will typically take effect instantly”; after that you can “request a rate limit increase”, with “no guarantees” (verified Oct 30, 2026). On Vertex pay-as-you-go, Google suggests “the global endpoint instead of a regional endpoint whenever possible”, a Quota Increase Request, smoothing traffic spikes, and ramping up gradually to avoid “acceleration limits” (verified Oct 30, 2026). Token prices: Gemini API pricing.

What about buying Gemini Code Assist seats for the bigger daily caps? Too late for new buyers: Google stopped selling new Standard and Enterprise subscriptions on October 9, 2026; existing ones keep working to the end of the contract (verified Oct 29, 2026). Details: Is Gemini Code Assist free?

Auth method, quota and fix at a glance

Auth methodQuota sourceFix
Google loginEnded June 18Switch auth
Free API key250/day, FlashWait or billing
Paid API keyProject tierBackoff, raise
Vertex AIShared quotaGlobal, QIR
Code AssistDaily per userWait for reset

From the quota page and rate-limits docs (Oct 30, 2026) and Code Assist docs (Oct 29, 2026), which also cap requests at “2 requests per second per user”.

Here’s what I’d do

If the error goes away after a minute, it was a per-minute limit: shrink what you send and keep going. If it sticks around on a free API key, you’re out of the day’s 250 Flash requests; wait for the midnight Pacific reset or link billing. If you’re still trying to sign in with a personal Google account or Google AI Pro, stop debugging the 429 and switch auth, because that path isn’t served anymore.

For heavier daily coding, compare the alternatives before paying: Gemini CLI vs Claude Code, Google AI Pro vs Ultra for Antigravity limits, and AI coding tools compared.

Using Antigravity instead of Gemini CLI? Antigravity’s “Individual quota reached” error is the sibling of this one.

Getting API error 400 Request contains an invalid argument instead of a 429? See the Gemini CLI 400 invalid-argument fix.

Getting a 503 instead of a 429? Gemini CLI “The model is overloaded” explains the server-side capacity error and the retry fix.

Getting a 403 “The caller does not have permission” instead of a 429? The Gemini CLI 403 fix goes by login type.

A 500 INTERNAL is Google’s server falling over, not a quota hit. Fixes in Gemini CLI Error 500.

FAQ

What does 429 RESOURCE_EXHAUSTED mean in Gemini CLI? It’s a rate-limit error: you exceeded a per-minute, daily or spend-based quota for your sign-in method, shown as “status”: “RESOURCE_EXHAUSTED” with reason rateLimitExceeded.

What does “Possible quota limitations in place or slow response times detected” mean? Older Gemini CLI versions showed it after a 429 to say they were switching the session to a Flash model; users reported the switch didn’t always happen.

How do I fix Gemini CLI 429 too many requests? Run /stats model, wait and retry with backoff, use a lighter model or prompt, switch to an API key or Vertex AI, or link billing for a higher tier.

When does the Gemini CLI quota reset? For Gemini API keys, daily quotas reset at midnight Pacific time; per-minute limits usually clear after a short wait; Google publishes no reset time for Vertex or Code Assist.

Does Google AI Pro raise Gemini CLI limits? No longer; since June 18, 2026 Google AI Pro and Ultra don’t serve Gemini CLI, and Google points those users to Antigravity CLI.

Will a new API key fix the 429? Not in the same project; Google says Gemini API rate limits apply per project, not per API key.

Sources

  1. Gemini CLI, Quotas and pricing (daily caps by auth method, per-minute note, pay-as-you-go, /stats model): https://geminicli.com/docs/resources/quota-and-pricing/ (verified 2026-10-30)
  2. Gemini API, Rate limits (RPM/TPM/RPD, per project, midnight Pacific reset, spend-based 429, tiers): https://ai.google.dev/gemini-api/docs/rate-limits (verified 2026-10-30)
  3. Google Cloud, Error code 429 (definition, messages, backoff, global endpoint, QIR): https://cloud.google.com/vertex-ai/generative-ai/docs/error-code-429 (verified 2026-10-30)
  4. gemini-cli issue #10722, “API Error: 429 over and over”: https://github.com/google-gemini/gemini-cli/issues/10722 (verified 2026-10-29)
  5. gemini-cli issue #10889, “Constant ‘RESOURCE_EXHAUSTED’ errors”: https://github.com/google-gemini/gemini-cli/issues/10889 (verified 2026-10-29)
  6. gemini-cli pull request #19786 (API-key fallback, closed unmerged): https://github.com/google-gemini/gemini-cli/pull/19786 (verified 2026-10-30)
  7. Gemini Code Assist consumer accounts (Login with Google removed June 18, 2026): https://codeassist.google/ (verified 2026-10-29)
  8. Gemini Code Assist sunset (no new Standard/Enterprise subscriptions from Oct 9, 2026): https://docs.cloud.google.com/gemini/docs/codeassist/sunset (verified 2026-10-29)
  9. Gemini Code Assist quotas (2 requests per second per user; daily caps): https://docs.cloud.google.com/gemini/docs/quotas (verified 2026-10-29)

Related reading