OpenAI Launches GPT-Live-1 API With Full-Duplex Voice Capabilities

OpenAI introduced the GPT-Live-1 API, a full-duplex voice model capable of simultaneous listening and speaking. The model improves handling of interruptions, pauses, and background noise, and is optimized for phone scenarios. Developers can customize tone and style, and integrate backend models like GPT-6 Astra for reasoning. Early testing on the Speak platform showed an 80% reduction in misinterpretations of learner pauses compared to traditional systems.