OpenAI makes voice more agentic with GPT-Live
OpenAIThe new voice stack listens and speaks continuously, while delegating harder work to frontier models in the background.
OpenAI introduced GPT-Live on July 8 as the next version of its voice experience in ChatGPT. The company frames it as a move away from turn-by-turn voice assistants toward something closer to continuous conversation.
What changed #
GPT-Live is built around full-duplex interaction: the system can listen and speak at the same time, decide when to pause, and handle interruptions more naturally. OpenAI says the model can keep the flow of conversation while delegating harder questions to a stronger frontier model in the background.
The rollout starts with GPT-Live-1 and GPT-Live-1 mini for ChatGPT users, with API availability planned later. OpenAI also says the new ChatGPT Voice experience can show visual cards during conversations for some topics.
Why it matters #
Voice is becoming another interface for agents, not just a nicer microphone for chat. If the delegation model works, people can ask for deeper work without breaking the rhythm of a live conversation. That matters for support, tutoring, hands-free work and any workflow where the user cannot stare at a screen the whole time.
Sources
- OpenAIopenai.com
Anthropic
Google