Error messages

Usage limit reached for all Pro models.

Usage limit reached for all Pro models.

Gemini CLI shows this in a boxed dialog when Google turns down a request and the CLI decides waiting a few seconds won't help:

Usage limit reached for all Pro models.
Access resets at 3:57 AM GMT+1.
/stats model for usage details
/model to switch models.
/auth to switch to API key.

That's the full form as 0.60.0 builds it. The reset time is borrowed from a December report on an older version, which said /stats for usage details on the third line. We read the code that builds it in Gemini CLI 0.60.0, the quota page in Google's own repo, and the GitHub issues where people post it.

Why it says all Pro models

The phrase isn't Google telling you that every Pro model has its own empty counter. The CLI swaps in "all Pro models" whenever the model that failed has "pro" in its name (a plain text check, unless an experimental setting is on). A Flash model gets named individually instead, as in Usage limit reached for gemini-3.5-flash.

The Access resets at line only appears when Google's response said how long to wait. The last line only appears if you signed in with a Google account, since that's the case where an API key is the alternative.

Which limit you hit

Gemini CLI sorts a quota refusal into two kinds. If Google's error names a per-minute limit, the CLI treats it as temporary and suggests retrying after 60 seconds. If it names a daily limit, that's terminal. The error underneath then reads You have exhausted your daily quota on this model., and the interactive CLI turns it into this dialog.

Google's quota page gives the daily ceilings as requests per user per day:

• Signing in with Google on Gemini Code Assist for individuals: 1,000
• Google AI Pro: 1,500, and Google AI Ultra: 2,000
• A free Gemini API key: 250, and Flash only
• Code Assist Standard: 1,500, and Enterprise: 2,000

The high-demand version is different

If Google's error mentions capacity (MODEL_CAPACITY_EXHAUSTED, or text like You have exhausted your capacity on this model.), the same dialog reads:

We are currently experiencing high demand for all Pro models.
We apologize and appreciate your patience.

That one is about Google's servers, and the choices differ to match: "Keep trying" is offered, alongside switching models and stopping. For a real usage limit there's no Keep trying. You get "Switch to" a fallback model, "Stop", and on a Google login below Ultra, "Upgrade for higher limits".

What to do

The quickest way on is Switch to, which moves the session to the fallback model the dialog names and keeps going. /stats model shows what's left per model and when each resets. If you need Pro specifically, you wait for the reset time or add capacity. Google's quota page points people who keep running out to pay-as-you-go through a Gemini API key or Vertex AI, which you can switch to with /auth.

When /stats says you have quota left

This is the part that sends people to GitHub. An August report on CLI 0.54.4, from someone on a Code Assist Standard license, showed the usage display at 1 to 8 percent while every request failed with the limit message. A September 20 report on 0.60.0 found Google's own quota endpoint saying 95.8 percent of Flash and all of Pro remained, and requests still came back 429 (the HTTP code for too many requests).

The same mismatch hit AI Pro users last winter. A collaborator on the repo wrote in January 2026 that they'd found "an underlying issue where usage limits were being enforced inconsistently between the backend and the CLI" and asked everyone to update. So step one is npm install -g @google/gemini-cli. One user who kept hitting the capacity version with quota to spare traced it, by their own testing, to heavy use of Google Search inside the CLI, which they suspected has a limit /stats doesn't show. If none of that explains it, the September issues are open and collecting logs.

For the Codex equivalent of a model being full, see “Selected model is at capacity. Please try a different model.”