getReasoningEffort

Reasoning-effort budget for reasoning models: default, none, minimal, low, medium, high, or xhigh (case-insensitive). default — the default — sends nothing, leaving the provider's own budget in force (medium on the OpenAI GPT-5 family). It is spelled out rather than left empty so the intent is readable in a settings file and in the generated API surface. Lower budgets cut latency and reasoning tokens at some cost in recognition quality; which values a given model accepts is decided by the provider, not by this SDK, so an unrecognized value is ignored rather than rejected.