OpenAI released two new voice models, GPT-Live-1 and GPT-Live-1 mini, on July 8, 2026, built on full-duplex technology that lets them listen and speak simultaneously for more natural conversations. GPT-Live-1 mini becomes the default replacement for Advanced Voice Mode for free ChatGPT users, while paid subscribers get the larger GPT-Live-1 model, which integrates with GPT-5.5 for deeper reasoning and supports live translation. The models can sustain conversations of 30-40 minutes, though a live-translation demo showed pronunciation and naturalness issues.
