The reasoning effort levels the GA Realtime API accepts for reasoning-capable realtime models
(gpt-realtime-2 / gpt-realtime-2.1 line). low is the provider default — it keeps latency down
for voice; raise only when task complexity justifies the added latency and reasoning tokens.
The reasoning effort levels the GA Realtime API accepts for reasoning-capable realtime models (gpt-realtime-2 / gpt-realtime-2.1 line).
lowis the provider default — it keeps latency down for voice; raise only when task complexity justifies the added latency and reasoning tokens.