OpenAI launches new “GPT-Live” voice models designed to handle real-time conversation by listening and speaking at the same time. The models support natural, interactive dialogue intended for low-latency voice use cases, enabling a more fluid back-and-forth between a user and the AI. OpenAI positions the release as an advancement over earlier voice interfaces by reducing the stop-and-start pattern often associated with traditional turn-taking systems. The rollout centers on enabling near real-time speech interaction, where the system can respond while continuing to process incoming speech. While details vary by coverage, both outlets describe the same core capability: simultaneous listening and speaking in a live voice setting. The announcements focus on the model behavior and user experience rather than on reported regulatory or safety outcomes. The company’s release indicates the models are built to make voice conversations feel more immediate and conversational, with the timing of responses closely tied to the ongoing spoken input.