AiReasoningLevel

Provider-neutral extended-thinking depth. off disables thinking where the model allows it.

Possible values:

• off
• low
• medium
• high
• max