OpenAI发布GPT-Live模型，任务完成率翻倍并开放API
OpenAI于2026年9月11日推出新一代语音模型GPT-Live。据发布信息，该模型的任务完成率较上一代翻倍，理解准确率达到92%，支持自然打断交互，已上线Genspark并开放API。
Telnyx作为发布合作伙伴，在官方推文中表示，GPT-Live为语音AI带来了更多细微差别，包括语气、时机、情感和意图的把握。Telnyx通过其Voice API以16kHz音频向模型传输语音，无需应用端重采样，使更多声音细节得以保留。
Telnyx还发布了关于使用GPT-Live与Telnyx Voice API进行外呼AI通话的指南。
- 2026-09-11 02:11 | X：RT Genspark: Congrats to the @OpenAI team on launching GPT-Live. 🎉 It sounds more natural. We ran it through 80 real restaurant booking calls first...阅读原文
RT GensparkCongrats to the @OpenAI team on launching GPT-Live. 🎉It sounds more natural. We ran it through 80 real restaurant booking calls first: task completion more than doubled over the previous generation, with 92% perfect comprehension.Interruptions are smoother. A quiet "mm-hmm" no longer sto...
- 2026-09-11 01:17 | X：RT Telnyx: GPT-Live from @OpenAI is here, and we're excited to be a launch partner. GPT-Live brings more nuance to voice AI: tone, timing, emotion, an...阅读原文
RT TelnyxGPT-Live from @OpenAI is here, and we're excited to be a launch partner.GPT-Live brings more nuance to voice AI: tone, timing, emotion, and intent. Telnyx carries 16 kHz audio to the model without app-side resampling, so more of the voice gets through.Read the guide on Outbound AI Calling w...