Agent model
Overview
The Agent Model setting picks the model Friday uses for its worker and dispatch agents.
Quick Start
Run /settings, choose Agent Model, and pick a model. The change applies to every task you start afterward.
Available models
| Option | Provider | Notes |
|---|---|---|
claude-sonnet-5 | Anthropic | The default |
gpt-5.6-terra | OpenAI |
Both drive worker and dispatch agents. Friday shows the active model below the input box.
Agent reasoning
A second setting, Agent Reasoning, sets how hard worker and dispatch agents think before they act. Run /settings, choose Agent Reasoning, and pick one of none, low, medium, high, xhigh, or max. The default is low.
Raise it for tangled work — a refactor that spans several files, a bug whose cause is unclear. Leave it low for routine edits, where the extra thinking costs time and buys little.
The setting covers worker and dispatch agents. It does not touch the planner model, which chooses its own effort per plan. Friday shows the active effort next to the model below the input box, as claude-sonnet-5 · low.
Planner model
A third setting, Planner Model, picks the model behind the Plan tool. Run /settings and choose Planner Model:
| Option | Provider | Notes |
|---|---|---|
gpt-5.6-sol | OpenAI | The default |
claude-fable-5 | Anthropic |
Agent Model and Planner Model are independent, and all four combinations work. Leaving Agent Model on claude-sonnet-5 and switching Planner Model to claude-fable-5 puts both on Anthropic; pairing claude-sonnet-5 with gpt-5.6-sol splits them. Planning happens before the work starts, so a change takes effect on your next multi-step prompt.
The Anthropic planner option moves from claude-fable-5 to claude-fable-5-1. Pick it the same way — the menu entry and everything else about the setting stay put.
Read next
- Settings reference for these keys and how they are stored.
- Subagents for setting a model per subagent.