OpenAI has launched GPT-Live-1 and GPT-Live-1 mini, new full-duplex models designed for more natural and intelligent conversational AI. These models represent a significant leap forward, allowing simultaneous speaking and listening for seamless interruptions and real-time features like live translation. GPT-Live-1 mini will become the default voice model for ChatGPT's Advanced Voice Mode, with the larger GPT-Live-1 available for paid subscribers. The previous architecture relied on separate speech-to-text, LLM, and text-to-speech components, which led to conversational limitations.The new models address issues like interrupting users and improving response intelligence. They integrate with advanced text models like GPT-5.5 to leverage robust search, reasoning, and agentic capabilities during ongoing conversations. This integration allows for more intelligent and contextually aware responses.During a demonstration, OpenAI showcased the models' ability to maintain silence and absorb conversation context until prompted. Furthermore, access to newer GPT models enables the voice mode to present information visually. This visual capability aligns with a broader industry trend among startups like Monogram, aiming to enhance assistant interactivity.OpenAI emphasizes that the new ChatGPT voice mode is built for extended conversations. Product lead Atty Eleti shared personal experiences of engaging in 30- to 40-minute discussions with the feature.