Models

Models and reasoning effort are set separately for three kinds of agent work: Captains, Threads and Subagents. Each setting is a model and, for models that support it, a Reasoning effort. Higher effort thinks longer, which takes more time and costs more.

Choose what each uses⁠

On the Agents card, three lines read like sentences:

• Captains use a model.
• Threads use a model.
• Subagents use a model.

Choose a line's chip and pick a model under Choose a model. Hover a model to choose its Reasoning effort. A model that does not support an effort says so: "This model doesn't support that reasoning effort."

Changes apply on the next run. Existing conversations adopt a changed default on their next run.

Three places to choose⁠

Your personal choice overrides the company default for the Thread and Subagent runs you send, and for your Personal Captain's model. A Space can override all three for everyone in it, and the Space wins where they differ. A model choice in a Captain's instructions or a file does not override any of them. See Agents and setup.

If you have not chosen, the line shows the company's choice.

Turn models on for the company⁠

Settings → Company → Agents also has a Models section: the catalog of models Whorl offers. Switch a model on in the Enabled column to make it available. Only enabled models can be chosen anywhere in the company. The table shows each model's Context size.

• A company keeps at least one model enabled, so the last one's switch is off-limits. Its tooltip says "An organization keeps at least one model enabled."
• A model Whorl no longer offers shows Unavailable.
• If the catalog cannot load, the page says "Whorl isn't offering any models yet." or "Whorl couldn't read the saved models. Reload to see them."

Only company owners and admins can change the catalog.

Related⁠

• Agents and setup
• Captains
• Usage and insights, which shows spend by model