Skip to main content
  • Instructions: write clear, detailed prompts for the agent. Check Prompting 101: How to Talk to Your Agents for more advice.
  • Advanced > Model selection: choose the model the agent runs on, and its reasoning effort.

​
Which model should I choose?

Start with an auto model, and use Standard unless you have a reason not to. The auto models page explains which one to use for different kinds of work. Pick a specific model when you need the agent to use that model rather than follow an auto-model list. Pinning a model does not guarantee identical responses. Auto models also handle fallback for new messages during detected provider outages. Pinned models do not switch automatically, even when unstable. See Provider outages and automatic fallback.

​
What reasoning effort should I set?

Reasoning effort only applies when you pick a specific model, and only the levels that model supports are selectable. See Reasoning effort for the levels and their tradeoffs.

​
Can I change the model without editing the agent?

Yes. Use the model picker in the composer. It applies to your next message in that conversation only and leaves the agent untouched.