/reasoning-effort
Sets the reasoning effort level of the current model (e.g. low/medium/high).
/reasoning-effort <value> sets the reasoning_effort API parameter for models that support adjustable reasoning (e.g. OpenAI's o-series/GPT-5 series). Depending on the model, a number or a level like low/medium/high is expected. On the command line this corresponds to --reasoning-effort VALUE; by default nothing is set (the model's own default applies).
✅ WHEN TO USE IT?
Force more thinking time for a difficult debugging task
/reasoning-effort high before a tricky bug hunt in concurrent code
Lower cost/latency for simple routine tasks
/reasoning-effort low before Aider only needs to adjust comments or formatting
⛔ WHEN NOT TO?
The current model doesn't support adjustable reasoning at all
Many models simply ignore the parameter, or the API rejects it.
Better: Check the model docs or /models to see whether reasoning_effort is supported; otherwise pick a different model
Wanting to always work with maximum reasoning by default
High reasoning effort increases latency and token cost even for tasks that don't need it.
Better: Set effort project-wide via config and only raise it per /reasoning-effort for individual hard tasks
SOURCES
- Aider Docs — In-chat Commands ↗ aider.chat