GPT-4o Realtime

GPT-4o Realtime is the full-capability member of OpenAI's realtime voice family, enabling stateful, interactive voice conversations with minimal latency. As a native speech-to-speech model it takes audio in and generates spoken responses directly, preserving GPT-4o's reasoning depth inside a streaming voice interface. It supports function calling for agentic workflows, and voice activity detection handles silence automatically to keep exchanges responsive. It suits sophisticated voice assistants and virtual agents that need to reason over conversation context, and applications where speech interaction complements or replaces text.