Some users saw quota being consumed during the free window. That’s usually because a subagent is still on GLM-5.3.
For fully zero quota usage, set all subagent models to GLM-5.3-Flash in Settings.
But use models based on the task: GLM-5.3 still works better for complex work.