Clarification on Daily Request Limit

[Zeor-Xain]

Gemini blocks usage even when it shows available quotas.
There is an inconsistency between the displayed limit and the actual usage.

Using Gemini CLI as a "free" user with a personal Gmail account for authentication.

The following limits are the results after one day of use with /model exclusively set to Auto (Gemini 3) /model (89% context left) until the model blocks.

The current state is a complete block of all models — PRO, FLASH, and LITE — refusing any prompt, whether /model is set to Auto (Gemini 3), Auto (Gemini 2.5), or any of the 5 manual options under Manual (gemini-3-pro-preview , gemini-3-flash-preview, gemini-2.5-pro, gemini-2.5-flash, 5. gemini-2.5-flash-lite).

Note that bellow /stats shows near-total consumption of 2.5 flash/lite, in addition to 1005 requests on 2.5-flash lite, which seems inconsistent since /model was set to Auto (Gemini 3) all day without being changed once.

Responses throughout the day were always prefixed with Responding with gemini-3-pro-preview, except for some very small tasks, such as making a commit. The quality of the responses seemed compatible with Gemini 3 Pro, with occasional interventions from Gemini 3 Flash.
Meanwhile, gemini-3-flash-preview 9 99.3% (Resets in 23h 11m) seems unused, indicating 99.3% free, which is inconsistent with the usage observed during the day.

But what really stands out is the indication of 48% free for the PRO models (2.5 and 3), even though they are completely blocked.

Why does /stats show 48% free for these models, as well as 99.3% free for gemini-3-flash-preview, when in reality all models are blocked?

│  Model Usage                 Reqs                  Usage left                                                                                  │
│  ────────────────────────────────────────────────────────────                                                                                  │
│  gemini-2.5-flash-lite       1005    6.4% (Resets in 18h 11m)                                                                                  │
│  gemini-3-pro-preview          44   48.0% (Resets in 20h 55m)                                                                                  │
│  gemini-2.5-pro                11   48.0% (Resets in 20h 55m)                                                                                  │
│  gemini-2.5-flash               6    6.4% (Resets in 18h 11m)                                                                                  │
│  gemini-3-flash-preview         9   99.3% (Resets in 23h 11m)                                                                                  │
│                                                                                                                                                │
│  Usage limits span all sessions and reset daily.                                                                                               │
│  /auth to upgrade or switch to API key.                                                                                                        │
│     

---

Is the persistence of this error message when typing any character in the prompt (appearing like an autocomplete), even after selecting stop, a bug?

  Responding with gemini-3-flash-preview
╭────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│                                                                                                                                                │
│ Usage limit reached for gemini-3-flash-preview.                                                                                                │
│ /stats for usage details                                                                                                                       │
│ /model to switch models.                                                                                                                       │
│ /auth to switch to API key.                                                                                                                    │
│                                                                                                                                                │
│                                                                                                                                                │
│   1. Keep trying                                                                                                                               │
│ ● 2. Stop                                                                                                                                      │
│                                                                                                                                                │
│                                                                                                                                                │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

[Zeor-Xain]

Related: #16766

[jackwotherspoon]

@Zeor-Xain do you happen to have prompt completion enabled by chance? Like where Gemini CLI suggests prompts as you are typing?

If you do, each suggestion is generated using a gemini-2.5-flash-lite call. This would be one case as to why you see the high number of flash lite requests.

Let me know if this is the case.

[Zeor-Xain]

Hello @jackwotherspoon ,

Thank you for your feedback.
Yes, the usage data in the OP represents usage with autocomplete enabled.

[jackwotherspoon]

then that would explain the 2.5 flash lite calls.

A user has open PR for disabling prompt auto-completion when quota for flash lite drops below 20% which should help reserve and save quota for core logic like model routing 👍

#15791

[jackwotherspoon]

But there may be another bug as to why 3 flash is showing as out of quota, will take a look