Reasoning effort
The five reasoning-effort levels, what each is suited for, and how to change one for a single turn.
Reasoning effort controls how much the model reasons before responding. It's set per message and applies to the next message you send; change it again for a different turn without changing anything else about the session.
| Level | Use it for |
|---|---|
| Low | Small, well-specified edits |
| Medium (default) | Ordinary work |
| High | Multi-file changes, tricky debugging, design decisions |
| Xhigh | Situations where a wrong answer is expensive |
| Max | The hardest problems: prefer waiting longer over iterating |
Choosing a level
Default to medium. Move up when the task has more ways to go wrong than a single obvious fix: multi-file refactors, debugging something non-deterministic, or picking between competing designs. Reserve max for cases where getting it right the first time matters more than getting an answer quickly.
Example
A one-line typo fix in a config file: low. Migrating a shared utility used across a dozen call sites: high. A production incident where a wrong fix would make things worse: xhigh or max.