OpenAI launches GPT-Live speech model that can speak and listen at the same time
OpenAI launched a new generation of voice AI model GPT-Live on Wednesday, aiming to achieve more natural voice conversations. The company says GPT-Live is built on a full-duplex architecture, meaning the AI can listen and speak at the same time, rather than waiting for a human to finish before responding. open artificial intelligence Official announcement.

GPT-5.6 release approved
The GPT-Live announcement comes a day after OpenAI said it would publicly release its flagship GPT-5.6 artificial intelligence model after the Trump administration requested a temporary delay in the release due to cybersecurity concerns.
Read here: Trump administration approves OpenAI’s GPT-5.6 rollout after security testing — here’s what’s next
OpenAI announced in an X post late Tuesday that it will publicly release its latest and most advanced AI model series, GPT-5.6, on Thursday. OpenAI CEO Sam Altman also confirmed the news in another article on X, writing: “Happy Architecture.” GPT-5.6 is OpenAI’s latest generation of AI models, including three versions: the flagship model Sol, as well as Terra and Luna.
OpenAI had earlier planned to publicly release GPT-5.6, but decided to delay the release at the request of the Trump administration. According to OpenAI, the U.S. government asked the company to delay the public release due to concerns about cybersecurity related to advanced artificial intelligence models.
How GPT-Live Voice works
After announcing the launch of GPT-5.6, OpenAI stated that GPT-Live’s full-duplex system allows users and artificial intelligence to speak simultaneously during a conversation. OpenAI says GPT-Live can show that it’s listening by giving small responses like “hmm” or “yes” when a person is speaking. GPT-Live can also engage in quick back-and-forth conversations without long pauses, the company added.
If a user asks a question that requires a web search, deeper reasoning, or more complex work, GPT-Live can offload that task to another senior person, the company said. you have a model Working in the background. OpenAI says GPT-Live continuously talks to the user while the background model handles requests.
OpenAI stated at the time of publication that GPT-Live uses GPT-5.5 as the backend model for these advanced tasks. The company added that GPT-Live will automatically use updated cutting-edge AI models when they are released in the future.
GPT-Live starts rolling out
OpenAI has begun rolling out two versions of the new model: GPT-Live-1 and GPT-Live-1 mini. According to the company, both versions will be released to Chat GPT Starting today, users around the world. OpenAI also said it plans to introduce GPT-Live to its API, allowing developers and enterprises to build applications using the new voice technology.