OpenAI unveiled GPT-Live, a new family of voice models designed to bring human-like chatbot interactions. The launch highlights a major shift from conventional voice assistants that wait for users to finish speaking before responding.
GPT-Live now allows AI to listen and speak simultaneously, bringing more natural, uninterrupted conversations. The company has begun rolling out two versions to users worldwide across iOS, Android, and the web.
OpenAI announced on its official blog, “We’re launching GPT‑Live, a new generation of voice models that make talking with AI feel much more like having a real conversation. GPT‑Live is built on a full-duplex architecture, meaning it can listen and speak simultaneously. GPT‑Live is also our smartest voice model yet. During conversations, GPT‑Live can show it’s paying attention with phrases like ‘mhmm’ or ‘yeah’, and engage in quick back-and-forth.”
OpenAI has redesigned how ChatGPT handles voice conversations. GPT-Live preserves the natural flow of conversation by offloading complex reasoning, web searches, and other advanced tasks to a larger backend model.
The company says this architecture improves efficiency while maintaining response quality. In addition, it offers features such as live translation. The model also provides subtle conversational cues such as brief acknowledgments, creating a smoother dialogue. OpenAI says these interactions reduce friction.
The company believes this architecture improves responsiveness without compromising the quality of answers. It will also enable the software to provide features such as real-time translation and enhanced visual responses for some questions.
GPT-Live is rolling out worldwide through ChatGPT. Subscribers with Go, Plus, and Pro plans will have access to the GPT-Live-1 feature by default. Free users can access GPT-Live-1 Mini. The launch builds on OpenAI's earlier release of audio models for developers this year.
Also Read: OpenAI Unveils Codex Micro, Marking its First Step into Consumer Hardware