Hourly ·
GPT-Live Breaks Voice Mode Out of the Dumb-Model Ghetto
OpenAI's new GPT-Live voice mode delegates complex questions to GPT-5.5 behind the scenes, ending years of voice users being stuck with outdated models. Early testers report hour-long conversations that actually stay useful.
For years, talking to ChatGPT felt like dialing into the future through a tin can. The voice was smooth, but the brain behind it was invariably a generation or two behind whatever model you could type to. That gap just closed.
OpenAI launched GPT-Live on Tuesday, a rebuilt voice mode that can delegate hard questions to GPT-5.5 in the background while keeping the conversation flowing in real time. The model can speak and listen simultaneously — a feature OpenAI says makes live translation and natural interruptions possible for the first time.
Early testers are impressed. Simon Willison, who had preview access for several weeks, reported carrying on hour-long brainstorming sessions while walking his dog. "The best feature is that it can delegate questions out to GPT-5.5 in the background, so you're no longer restricted to a voice model that's several years behind the frontier," he wrote on Hacker News.
The release comes just days after the Trump administration cleared OpenAI's GPT-5.6 suite for public launch, and two weeks after Anthropic shipped Claude Cowork on mobile. The voice AI race is accelerating on all fronts.
Not everything is seamless. Willison noted a bug where the model interrupted him mid-sentence to laugh at jokes he wasn't making — behavior OpenAI has since dialed back. Other early users described the model as still too eager to tack friendly sign-offs onto responses that should have ended with "ok."
The deeper tension threading through the launch is whether people want an AI that sounds like a colleague or one that sounds like a utility. Several commenters on Hacker News said they'd prefer the Star Trek computer over the friend simulator. OpenAI seems to be hedging: GPT-Live stays conversational but now has the horsepower to be useful.
GPT-Live is rolling out starting today across ChatGPT apps on iOS, Android, and the web.
Sources: OpenAI, TechCrunch, Hacker News
GPT-live突破了 dumb-model的 ghetto_voice模式
阿里云Qwen的简短翻译:阿里云Qwen的新GPT-Live语音模式将复杂的问答委托给幕后使[K 用的GPT-5.5,结束了语音用户的年代久远模型。早期测试者报告说进行了一小时以上[K 的对话并且仍然保持有用。
← 周报 Hourly · 2026-07-08 20:00 UTC GPT-Live 开启语音模式走出低效模型的困境[K OpenAI最新推出的GPT-Live语音模式在后台委派复杂的问答任务给GPT-5.5,结束了数[K 年来的语音用户只能使用过时模型的情况。早期测试者报告称,能够保持有用性的对话[K 时间长达一个小时。多年以来,与ChatGPT交谈就像是通过一部穿越未来的电话拨号机[K 。
More Hourlies Stories
Content on Anagnorisis is summarized, paraphrased, and editorialized from publicly available sources for length and clarity. Original sources are linked where available. All trademarks belong to their respective owners.
