Seller Home has learned from foreign media reports that OpenAI has introduced the GPT-Live-1 model into its API, supporting full-duplex real-time voice interaction, interruption handling, and other features. It integrates speech understanding and output, reduces latency, and can connect to tools for enterprise system operations. Early evaluations show that after Speak integrated it, false interruptions dropped by nearly 80%. In tests, the model outperformed GPT-Realtime-2.1, and the front-end speech layer is priced at $0.05 per minute.

The technical highlights behind this are notable, with leading test rankings and multiple new voice options. In the future, it plans to add more voice and language support, which is expected to expand its market influence and drive the development of voice interaction technology.