Deepgram has introduced a new reasoning_mode parameter for its Voice Agent, enabling developers to control the reasoning effort of supported OpenAI reasoning models. This addition directly maps to OpenAI's reasoning_effort configuration option.

What Changed

The Voice Agent's think provider configuration now accepts a reasoning_mode parameter with three options:

• low: Minimal reasoning, faster execution
• medium: Balanced reasoning and performance
• high: Maximum reasoning depth for complex tasks

Implementation

Configure reasoning mode directly in your Voice Agent settings:

{
  "type": "Settings",
  "agent": {
    "think": {
      "provider": {
        "type": "open_ai",
        "model": "gpt-5",
        "reasoning_mode": "medium"
      }
    }
  }
}

Next Steps

Developers using OpenAI reasoning models with Deepgram's Voice Agent should update their provider configurations to leverage this new parameter. This allows for granular optimization of reasoning behavior based on specific use cases and performance requirements. See the Configure the Voice Agent documentation for additional details.