Thinking Level
Higher levels think longer and use more tokens.
Last updated on
Default: Balanced · Type: Quick, Balanced, Thorough
Sets how much the assistant reasons before it answers.
Quick replies fastest and uses the fewest tokens. Good for simple lookups and small edits to a query you already have.
Balanced is the default and suits most reports.
Thorough takes longer and uses more tokens, because the thinking counts towards your usage even though you do not see it. Use it for complex reports with many joins, or when a Balanced answer was wrong.
The level applies from your next message. You can change it part way through a conversation. The first reply after a change may use more tokens, because the earlier messages may need to be read again at full cost rather than reused.
How it looks
Thinking Level
Higher levels think longer and use more tokens.
Where to change it
- AI Settings in your profile menu, the page with every AI setting.
- Quick AI settings in the chat input, without leaving your conversation.
See all AI settings.