OpenAI has launched GPT-Live, a new generation of voice models that can listen and speak simultaneously — making conversations with AI feel dramatically more natural than previous systems.
Announced on July 8, GPT-Live introduces a full-duplex architecture that processes input and generates output continuously, rather than in discrete turns. The result is a voice experience where the AI can acknowledge what you're saying with phrases like "mhmm" or "yeah," engage in rapid back-and-forth, or stay quiet when you need a moment to think.
Smarter Voice, Seamless Delegation
GPT-Live is OpenAI's smartest voice model to date. When a question requires web search, deeper reasoning, or complex work, it delegates to the latest frontier model behind the scenes — currently GPT-5.5 — while continuing to maintain the conversation flow. This architectural decoupling means users get natural interaction speed without sacrificing intelligence.
Two versions are rolling out: GPT-Live-1 (default for ChatGPT Plus, Pro, and Go subscribers) and GPT-Live-1 mini (default for Free users). Both are available globally on iOS, Android, and web.
Beyond Cascaded and Turn-Based Models
Previous voice AI systems operated in one of two ways. Cascaded systems (like the original ChatGPT Voice) chained three separate models together, causing information loss and stilted responses. Turn-based models (like Advanced Voice Mode) improved latency but still required the AI to wait for silence before responding, leading to unnatural interruptions.
GPT-Live overcomes both limitations by continuously processing audio, making interaction decisions many times per second — whether to speak, listen, pause, interrupt, or invoke a tool. The model also has a better sense of timing and can even perform live translation.
Safety by Design
OpenAI expanded safety testing specifically for voice interactions, including evaluations for self-harm, emotional reliance, violence, and sexual content. The system includes real-time safeguards that can steer the model toward safer responses or end conversations in high-risk cases. Dedicated protections for teen users are also built in, with parental control options.
Availability
GPT-Live is rolling out now to ChatGPT users globally. OpenAI plans to bring the models to its API soon, with a sign-up form available for developers and enterprises. Legacy voice modes remain accessible for users who prefer them.
The launch marks a significant step toward OpenAI's vision of "truly natural human-AI interaction" — where collaborating with AI feels as fluid as working with another person.
