OpenAI releases GPT-Live voice model, making ChatGPT more like real human conversation

CoinFeed July 9 news, according to Global Market Broadcast reports, OpenAI officially launched the next-generation voice models GPT-Live-1 and GPT-Live-1 mini today, comprehensively upgrading ChatGPT's voice capabilities. The new models adopt a 'full-duplex' architecture, capable of listening and speaking simultaneously, allowing users to interrupt at any time during AI responses, making interactions closer to real conversation. Another improvement of the new model is the separation of the voice interaction layer from the reasoning layer. When questions require web search or more complex reasoning, GPT-Live hands off tasks to the backend GPT-5.5 processing while maintaining smooth conversation flow. Paid users can choose from three reasoning levels: 'Instant', 'Medium', and 'High', balancing speed and intelligence according to needs. Additionally, GPT-Live supports real-time translation, displaying weather and stock information via visual cards, etc.