OpenAI rolls out GPT-Live-1 and GPT-Live-1 mini globally, introducing full-duplex architecture that enables ChatGPT to listen and speak at the same time — with background task delegation to GPT-5.5 for complex reasoning, marking voice AI's shift from turn-based chat to continuous conversation.
On July 8, 2026, OpenAI launched GPT-Live, a new generation of voice models that fundamentally changes how ChatGPT conversations work. Unlike the previous Advanced Voice Mode, which required turn-taking, GPT-Live uses full-duplex architecture — the model can listen and speak simultaneously, insert natural acknowledgments like 'mhmm' or 'got it,' pause when the user hesitates, and handle interruptions gracefully. Two versions were released: GPT-Live-1 for Go, Plus, and Pro subscribers, and GPT-Live-1 mini for free-tier users, both rolling out globally on iOS, Android, and the web.
The most significant architectural innovation is task delegation. When a query requires web search, deep reasoning, or multi-step agentic work, GPT-Live hands the heavy lifting to GPT-5.5 running silently in the background while maintaining the voice conversation. Once GPT-5.5 completes the task, the result is woven back into the voice stream. Users can choose from three reasoning levels — Instant, Medium, or High — each corresponding to a different GPT-5.5 model tier. The system also introduces visual cards for structured information like weather and stocks, supports nine remastered voices, and works with Apple CarPlay for in-vehicle use.
The launch comes as voice AI competition intensifies across the industry. Google's Gemini Live, ByteDance's Doubao, and xAI's Grok Voice have all made aggressive moves in recent months, and ElevenLabs crossed $500 million in annual recurring revenue. OpenAI's counter is ecosystem integration: GPT-Live is deeply coupled with ChatGPT's existing features including memory, search, image upload, and file analysis. With over 150 million people already using ChatGPT voice and dictation features weekly, the upgrade instantly reaches one of the largest voice AI user bases in the world.
Safety measures were prominently featured in the launch. GPT-Live includes real-time content classifiers that can steer conversations toward safer responses or end sessions in high-risk situations. Teen-specific protections include age-appropriate behavior training and parental notifications. The model only uses preset voices and is designed to refuse impersonation requests — a direct response to the regulatory environment shaped by the EU AI Act's deepfake rules and the FTC's TAKE IT DOWN Act enforcement. API access is planned but not yet available at launch, and video and screen sharing remain on the legacy voice mode for now.
