You can’t directly set reasoning_effort from Vapi yet when using Azure AI models the standard way. Vapi currently just references the deployed Azure model name, not a custom Azure agent/config where reasoning settings are defined. The workaround is usually to route through your own middleware/proxy and inject the reasoning_effort parameter (low, medium, high, etc.) into the Azure OpenAI request before it reaches Azure. Azure itself supports the parameter, but Vapi doesn’t fully expose it in the UI/config currently. @campaholic