🎙 OpenAI Introduces GPT-Live with Full-Duplex Architecture
The new generation of voice models allows ChatGPT to listen and speak simultaneously. This technology ensures natural dialogue: the model uses backchanneling phrases ("mhm", "uh-huh") and reacts instantly to user interruptions. For complex tasks, GPT-Live can delegate queries to GPT-5.5 level models.
🌍 The transition to full-duplex architecture marks a leap from turn-based messaging to real-time streaming interaction, which is critical for creating fully functional voice agents.
👤 The voice mode will become significantly more "human" and responsive, minimizing waiting pauses and allowing for live dialogue without the need to press buttons.
Source 1: https://openai.com/index/introducing-gpt-live/