Pick which model answers your prompts
getting started
// what it does
-m (or --model) sets the model for one session and accepts either an alias — auto, pro, or flash — or a full name like gemini-2.5-pro. The default is auto, which routes each turn between a Pro and a Flash model by task complexity. Set GEMINI_MODEL in your environment to change the default permanently, or run /model inside a session to switch via a dialog.
// shell
$ gemini -m flash$ gemini -m gemini-2.5-pro$ export GEMINI_MODEL=gemini-2.5-flash// gotcha
auto can silently downgrade to Flash mid-task to save quota, which shows up as a sudden drop in answer quality. Pin -m gemini-2.5-pro (or -m pro) when you need consistent reasoning and do not care about speed.