Model and effort

Every session runs on one model at one reasoning effort. Together they set how thorough or how fast the session is.

Model

/model switches the session between three models: Fable, Opus, and Sonnet.

Fable is the recommended choice for deal work: it's the most thorough, and it often takes longer to respond. Sonnet is the fastest; use it when you want quick answers more than depth. Opus is the middle option: use it when Fable's pauses are too long for the task but you want more depth than Sonnet.

The choice is per-session: it holds for the rest of the current session. New sessions open on Fable until an admin picks a different Default model in Workspace settings.

Effort

/effort opens a picker of four levels, which set how much reasoning the model puts into each response: High, Extra high, Max, and Ultra. High is the balance struck for quick turnaround. Extra high and Max buy deeper reasoning for heavy modeling and the most complex analysis, and slower responses with them. Ultra is for the most exhaustive work and takes the longest.

The level a session opens on follows its model: High on Fable and Opus, Extra high on Sonnet. An admin can set a different Default reasoning effort for the organization in Workspace settings; Ultra stays a per-session choice. Switching model mid-session leaves the effort where it is.

To show or hide the terminal's thinking display, use /show-thinking.

A second opinion

The terminal can consult a senior reviewer model about its own work. While it does, the conversation reads "Asking the VP", and the status line names the reviewer.

/ask-vp turns those consultations off or back on for the current session. Run it on its own to pick between the reviewer and "No VP", or type /ask-vp off and /ask-vp on to switch it directly. An admin sets whether the organization has Ask the VP at all, and which model reviews, in Workspace settings.

Learn more

• Workspace settings: the model, effort, and reviewer new sessions open on, set by an admin.
• While the terminal works: what long pauses look like and when to worry.
• Model provider: the choice, for the whole organization, of which service processes model requests, set by an admin.