How hard the model thinks (/reasoning)
| Command | Does |
|---|---|
/reasoning | shows the current chat's thinking effort |
/reasoning high (or off, low, medium, xhigh) | sets it for this chat |
/reasoning high --default | also makes it the default for new chats |
- Default:
high, since coding and long multi-step work gain the most from deeper thinking. - Turn it down (
medium,low, oroff) for quick questions or to save money. More thinking costs more tokens and takes longer per step. - Every reasoning model. It applies to every model that supports reasoning. When the model you're on doesn't, Houston says so plainly, since a setting that silently does nothing looks like a bug.
- Desktop app: the same setting is in the reasoning menu next to the model picker. Changing it while the agent works applies from its next step.