GPT-Live: ChatGPT Listens & Talks Simultaneously

OpenAI's new GPT-Live lets ChatGPT listen and talk at the same time. This is a significant change because previous versions operated in discrete turns, often interrupting users or losing context. GPT-Live continuously processes incoming audio while generating its own output. It decides many times per second whether to speak, listen, pause, or use a tool. This means the model no longer waits for a turn, and it can interject with acknowledgments like "mhmm" or handle interruptions seamlessly. It can even perform live translation. When a question requires deeper thought or a web search, GPT-Live hands the task to GPT-5.5 in the background, continuing the conversation until the answer is ready. Users can also choose a reasoning level, from "Instant" for quick replies to "Medium" and "High" for more complex questions. ChatGPT Voice is also gaining visual answers, displaying on-screen cards for things like weather or stock information alongside spoken replies. With over 150 million people already using ChatGPT Voice and Dictation weekly, this upgrade impacts a large audience. Safety measures have been reworked for the new format, with real-time safeguards in place. The system can steer conversations toward safer responses, provide additional safety messages, or even end a session in high-risk situations, including offering crisis helpline support for self-harm discussions. This matters because it creates a more natural, responsive, and safer interactive experience with AI.