New LLM Model Support

Deepgram has extended the Voice Agent API with support for three new language models, enabling developers to leverage the latest AI capabilities for voice-based applications:

• OpenAI GPT-5.3 Instant (gpt-5.3-chat-latest) - latest lightweight variant for efficient inference
• OpenAI GPT-5.4 (gpt-5.4) - newest flagship model from OpenAI
• Google Gemini 3.1 Flash Lite (gemini-3.1-flash-lite) - optimized Gemini variant for faster responses

Integration & Usage

Developers can specify these models in their Voice Agent configuration by setting the provider.model field in the agent's think settings. For example, you can now configure GPT-5.3 Instant as follows:

{
  "type": "Settings",
  "agent": {
    "think": {
      "provider": {
        "type": "open_ai",
        "model": "gpt-5.3-chat-latest"
      }
    }
  }
}

Bug Fix

The update also addresses a critical issue where the GPT-5.2 Instant model was using an incorrect model ID and pricing tier. The model now correctly uses the ID gpt-5.2-chat-latest and is properly assigned to the Advanced pricing tier. Existing implementations should verify their configurations if they're using GPT-5.2.

For complete documentation on all available models and their associated pricing tiers, refer to the Voice Agent LLM Models documentation.