Pick any harness or model per thread. When you start a new thread, choose the harness and the model it runs — starting with Anthropic and OpenAI models, with more on the way.
Set a default for each Space. Every Space can have its own default harness and model. Give Engineering a heavyweight coding setup and Support a fast, efficient one — new threads in each Space start with the right configuration automatically.
Choose models for automations. Select the harness and model each automation runs on, so recurring reports and monitoring use an efficient model while high-stakes work gets a more capable one. It's the simplest way to manage your automation costs.
Tune reasoning level and speed. Everywhere you pick a model — threads, Spaces, and automations — you can also set the reasoning level and speed, trading depth for responsiveness depending on the task.