Reasoning effort and Fast mode
Set how hard the model thinks before it acts, what each effort level costs, and when Fast mode is available.
On this page
Reasoning effort
Effort sets how hard the model thinks before it acts. Deeper thinking gives better results on hard tasks, but it is slower and uses more quota. Click the Effort button beside the message box to change it.
| Level | Good for |
|---|---|
| None, Minimal, Low | Quick, simple edits where speed matters. |
| Medium | Most everyday work. |
| High, XHigh, Max | Hard logic, debugging, architecture and large refactors. |
| Adaptive | The model decides how deeply to think for each step. |
What you can choose depends on the model:
- Each model offers only the levels it supports.
- Some models have no effort control at all. Then the button says so, and the model runs at its own fixed depth.
- If the button says set by the model, the model's own default is used.
RoGenie remembers the effort you pick for each model.
Higher effort can be cheaper overall on hard problems, because the model may get it right in fewer turns. On simple tasks it just costs more. Start at the default and raise it when RoGenie struggles.
Advanced thinking options for Claude and Gemini models (extended thinking, adaptive thinking, the thinking budget ratio and thought summaries) are in Agent settings.
Fast mode
Fast mode gives up to 2.5× faster output at 6× the price. The Fast button appears beside the message box only when the selected model and route support it (currently Claude Opus 4.6 on a direct Anthropic route, such as your own Anthropic key). Click it to turn Fast mode on or off; RoGenie confirms with "Fast Mode enabled (6× pricing)".
The 6× is Anthropic's price for fast output. On your own Anthropic key, Anthropic charges it to your account and no RoGenie quota is used, so use Fast mode only when speed matters more than cost.