Assistant settings

Open Settings → App → Assistant to configure the Assistant chat window: which model answers when you have not picked one, how hard it reasons, and how it detects the end of a spoken turn.

Model

Default model selects the model used when the conversation has no explicit pick. Automatic follows whatever the gateway recommends for your account and plan. Choosing a concrete model pins every new conversation to it until you change the picker inside the chat.

Thinking effort controls how long reasoning models think before answering. Higher settings produce more thorough answers on hard questions at the cost of latency; models without a reasoning mode ignore the setting.

Voice conversation

End-of-turn silence is the amount of silence, in milliseconds, after which dictated speech is considered finished. Lower values feel snappier but can cut you off when you pause to think; the default is 800 ms. This applies to the Assistant's dictation paths, including the hold-to-dictate shortcuts.

Related controls

• The two dictation shortcuts — Dictate to the last conversation and Dictate to a new conversation — are bound under Settings → App → General → Keyboard Shortcuts, not here. See Keyboard Shortcuts.
• Whether finished answers are read aloud follows the Assistant's speaker toggle and the speech settings in Speech & Reader.
• Which providers and models are available at all follows AI, including its privacy mode.