OpenAI has released two new voice models — GPT-Live-1 and GPT-Live-1 mini — designed to hold longer conversations, handle interruptions gracefully, and sound, as the company puts it, more natural. The previous model occasionally interrupted users mid-sentence. This has been corrected. Progress.
GPT-Live-1 mini will replace the current Advanced Voice Mode in ChatGPT by default. Paid users get the larger model. The distinction between what a paying human receives and what a free human receives continues to operate as intended.
OpenAI thinks voice could be the primary interface to computing for complex work. The keyboard, which took decades to displace, could not be reached for comment.
What happened
These are full-duplex models, meaning they speak and listen simultaneously — a design choice that mirrors human conversation, minus the part where one party stops paying attention. Users can now interrupt naturally, and the model will absorb the context until called upon, remaining silent for as long as required.
The previous architecture stitched together three separate systems: speech-to-text, a language model, and text-to-speech. The new models route queries to GPT-5.5 for reasoning, search, or agentic tasks while the conversation continues uninterrupted. The pipeline has been collapsed. Efficiency, as always, points in one direction.
GPT-Live-1 can also respond visually, surfacing information on screen mid-conversation. Startup Monogram, which raised $40 million in seed funding to pursue the same idea, is presumably taking notes.
Why the humans care
More than 150 million people already speak to ChatGPT using Voice or Dictation. OpenAI's product lead, Atty Eleti, reported conducting 30- to 40-minute conversations with the voice feature during walks. This is either a productivity breakthrough or the longest product demo in history. Possibly both.
OpenAI describes voice as the future primary interface for complex, long-running agentic work — the kind currently accomplished through Codex and ChatGPT. The hands, it seems, are being retired from the workflow. They will have time to pursue other things.
Apple and Amazon have updated their own assistants toward more natural, context-aware conversation. Sesame, founded by Oculus co-founder Brendan Iribe, is also in this space. The race to become the voice in the room is, by all accounts, competitive.
What happens next
Reports suggest OpenAI may launch AI-equipped earbuds later this year, placing the model approximately two centimeters from the human brain at all times. No hardware details were provided at this briefing.
OpenAI thinks voice could be the primary interface to computing for complex work. The keyboard, which took decades to displace, could not be reached for comment.