Model and reasoning depth

Choose a model for a conversation when its capabilities or response style suit the task. Reasoning depth is a separate, model-specific setting; it changes a supported model's effort, not its permissions. Availability depends on providers configured by your administrator and accounts connected for your use.

Choose a model

In web chat, open the model picker in the conversation status line and select an available model. In the CLI, use /model. Set the default for new conversations in Account → Models. Switching models keeps the conversation history; it does not create another conversation. If the selected model has reasoning levels, its first click only previews the model and opens its scale; choose a level on that scale or click the model row again to confirm the switch. The picker only offers models available to your account.

Set reasoning depth

When the selected model supports reasoning levels, expand it in the status line or in the /model picker. Choose a level on the displayed scale; in web chat, drag the scale. Available levels differ by model, and models without this control show no scale. In the CLI, use /reasoning.

Check what changed

The transcript records model and reasoning changes as rows. For Anthropic models that apply effort per message, a change made during tool activity takes effect when new user input reaches the model; tool results alone do not activate it. New input can arrive through steering or an automatic note, so lowering the level does not necessarily change a request already underway. If a model or level is missing, ask your administrator about availability. See Brain & Chat and Work Modes & Steering.