OpenAI Launches GPT-Live, Full-Duplex Voice Model for Natural Conversation in ChatGPT.

OpenAI announced on LinkedIn the launch of GPT-Live, a new generation of voice models designed to make AI conversations feel more natural and responsive. The update is rolling out immediately to all ChatGPT users on iOS, Android, and web, with API access coming soon.

GPT-Live represents a significant architectural shift from previous voice models. Built on full-duplex architecture, it can listen and speak simultaneously rather than taking turns. This enables more fluid back-and-forth exchanges, better temporal awareness during conversations, and the ability to perform live translation while maintaining natural dialogue flow.

The full-duplex design allows GPT-Live to interrupt itself, respond to conversational cues in real time, and maintain context without the awkward pauses that characterized earlier voice interfaces. According to OpenAI, the model also has a more defined personality, making it a better conversationalist for extended interactions.

For queries requiring deeper analysis, web search, or complex problem-solving, GPT-Live can delegate to OpenAI's latest frontier models behind the scenes. The system continues the conversation naturally while processing occurs in the background, then integrates results when ready. This approach separates conversational fluency from computational depth, letting each model handle what it does best.

The delegation architecture mirrors a pattern seen in other AI voice products, where lightweight models handle interaction while heavier models tackle reasoning. OpenAI has not specified which frontier model GPT-Live delegates to, though the company recently previewed GPT-5.6 variants in limited testing.

Users can access GPT-Live by tapping the Voice button in ChatGPT. No additional setup or subscription tier is required. The rollout began today and will reach all users across mobile and web platforms over the coming days.

OpenAI stated API access is coming soon but provided no specific timeline. Developers building voice-enabled applications with Claude, Gemini, or other AI models will likely watch for API pricing and latency benchmarks once GPT-Live becomes available for integration.

The launch continues OpenAI's push to make ChatGPT more versatile beyond text-based interactions. Voice has become a key competitive front as AI assistants move into productivity workflows, customer service, and creative collaboration. Full-duplex conversation, if it delivers on the promised naturalness, could set a new baseline for voice AI quality.

For small business owners and builders using ChatGPT in customer-facing applications or internal tools, GPT-Live may reduce the friction of voice interfaces. Live translation capabilities could also open workflows for multilingual support without switching between tools. OpenAI has not detailed which languages are supported at launch or how translation quality compares to dedicated services.