Reasoning effort

Reasoning effort lets you choose how much reasoning the current model should use. Available levels and their effects depend on the model; a higher level may take longer and does not guarantee a better answer.

Choose a level

1. Open /models and select a model.
2. Enter /efforts and select one of the levels shown.
3. Check the confirmed value beside the model in the footer; it is also shown in the diagnostic sidebar.

This model offers low, high, and max, with high labeled as the default. The cursor highlights low, which alone does not mean it has been applied. Confirm the selection and then check the current value. Other models may offer different levels.

The effort alias helps find /efforts. Levels come from the current model catalog: not every model offers low, medium, and high. If no levels are available, use the model's normal behavior or choose another supported model.

The selection is stored for the current session. When switching models, Mogick keeps it only if the new model supports it; otherwise it uses the new model's supported default or clears the selection. Check the value after switching or resuming.

Effort versus model configuration

/efforts changes the session's reasoning_effort. It is separate from models[].thinking.enabled in model configuration.

For custom models, thinking.enabled is a per-model setting: true requests enabled thinking, false requests disabled thinking, and omission leaves the request field unset. The provider must support that setting. In 26.9.11, budget_tokens is retained in configuration but does not change the outbound request; do not rely on it to limit reasoning usage.

There is no --effort CLI flag. See Authentication and models and Configuration reference.