gemini-cli - 💡(How to fix) Fix Quota flapping (9% to Limit Reached) on Gemini Code Assist Standard with Gemini 3 [1 comments, 2 participants]

What happened?

While using the "Auto (Gemini 3)" (gemini-3.1-pro-preview) model on a Gemini Code Assist Standard plan, the quota display is extremely unstable.

1. It shows "9% used" (consistent with the 1,500/day plan limit).
2. Without significant usage, it suddenly changes to "limit reached," blocking all agent actions.
3. After some time, it reverts to "9% used" again. This "flapping" behavior suggests a conflict between the Standard plan's high quota and a lower, hidden cap for the preview model.

What did you expect to happen?

I expected to be able to use the full 1,500 requests per day granted by my Standard subscription, or at least have a consistent and predictable daily limit for the Gemini 3 model that doesn't fluctuate or block the session unexpectedly.

Client information

• CLI Version: 0.39.1
• Git Commit: 4d73f3413
• Session ID: d9059329-205e-4f5f-85ce-3b80943d42bf
• Operating System: win32 v24.15.0
• Sandbox Environment: no sandbox
• Model Version: auto-gemini-3
• Auth Type: oauth-personal
• Memory Usage: 322.2 MB
• Terminal Name: xterm.js(6.1.0-beta.197)
• Terminal Background: #191a1b
• Kitty Keyboard Protocol: Supported

Login information

Logged in via Google Account with a Gemini Code Assist Standard license assigned in Google Cloud Console.

Anything else we need to know?

• CLI Version: v0.39.1
• Model in use: Auto (Gemini 3) / gemini-3.1-pro-preview
• Context: The issue appears even when the total daily consumption is very low (e.g., less than 150 requests total). This prevents professional use of the high-reasoning Gemini 3 model for long-form creative writing and complex code analysis.