Technology and Science · 08 July, 2026 · 2 min read
OpenAI Launches GPT-Live Voice Models With Full-Duplex Listening And Speaking
GPT-Live-1 and GPT-Live-1 mini enable full-duplex listening and speaking. Global rollout to all ChatGPT users, including free tier.
Depth of coverage on API availability (soon vs delayed/gated).
9 of 11 outlets skipped it: GPT-Live can show visual cards during voice..
15outlets compared
Same story, two versions
tap a side to read it in full
MediaPost
“will accelerate the value of intent data available to advertisers across the OpenAI Ads platform”Read the original ↗
Crypto Briefing
“OpenAI framed the release as a step toward more fluid human AI interaction.”Read the original ↗
MediaPost emphasises ad/intent angles; Crypto Briefing stresses humanlike UX and safety controls.
GPT-Live Goes Full-Duplex
OpenAI launched GPT-Live, a family of voice models optimized to process spoken instructions, ahead of a broader GPT-5.6 release.
“OpenAI Group PBC today introduced GPT-Live”
SiliconANGLE says GPT-Live is designed to power ChatGPT’s voice mode and that GPT-Live includes two models on launch: GPT-Live-1 and GPT-Live-1-mini.
Reuters, via SRN News, reports that GPT-Live can listen and speak simultaneously in real time, and that OpenAI will roll out two versions—GPT-Live-1 and GPT-Live-1 mini—to users globally on Wednesday.
SiliconANGLE adds that GPT-Live features a full-duplex architecture and that, during conversations, the model regularly checks whether it should interrupt the user, search the web or perform some other task.
In the same launch framing, SiliconANGLE says GPT-Live-1 scored 75.5 in OpenAI’s evaluations and that, when GPT-Live receives complex prompts, it routes them to GPT-5.5.
Voice Mode, Real-Time Translation
MediaPost describes GPT-Live as built on a full-duplex architecture that lets the voice model listen and speak at the same time, and it says GPT-Live can pay attention with phrases such as “mhmm” or “yeah.”M
In a demo described by MediaPost, a user asks, "I'm making a sweater for my grandson, but I don't want the needles to be too big, what do you think?" and GPT-Live replies, "If you go up [a size], it will get very baggy,".M
SiliconANGLE says ChatGPT’s voice mode previously had to wait until users finished speaking before generating a response, which meant it wasn’t capable of performing real-time translation.
SiliconANGLE also says GPT-Live removes the need for the model to wait until the user finishes describing a request, because it can interrupt, search the web, or perform other tasks while the conversation continues.
Business Insider adds that OpenAI says the assistant can listen and speak simultaneously so it can translate in real time rather than waiting for the speaker to finish.
Rollout, API Plans, and Safety
SiliconANGLE says GPT-Live is rolling out to the iOS, Android and web versions of ChatGPT with support for multiple languages, with GPT-Live-1 set as the default option in paid ChatGPT plans and GPT-Live-1-mini powering the free tier.
“make it available to developers via an application programming interface”
SRN News reports that OpenAI said it will roll out two versions of GPT-Live to users globally on Wednesday, and it frames the launch as part of OpenAI’s push for voice-based software agents that can complete tasks in real time.
SiliconANGLE states that OpenAI plans to make GPT-Live available to developers via an application programming interface, and it says GPT-Live can use GPT-5.5 to browse the web, explain complex topics and perform a range of other tasks.
SQ Magazine says OpenAI built new audio-native safety evaluations and red-teamed GPT-Live for risks specific to voice, including self-harm, psychosis and mania, emotional reliance on AI, violence, and sexual content.
In the same safety framing, SQ Magazine says parents can restrict teen access to ChatGPT Voice through Parental Controls and that linked parents may be notified in higher-risk situations involving signs of self-harm or suicidal intent.