Term
GPT-Live-1
OpenAI's real-time voice model, the default voice experience in ChatGPT since 8 July 2026. It listens and speaks at the same time (bidirectional), allows interruptions, and hands complex questions off to a stronger text model. Proprietary access via ChatGPT.
GPT-Live-1 — explained in more detail
GPT-Live-1 is OpenAI’s real-time voice model for ChatGPT. It became the default voice experience on 8 July 2026, replacing the previous “Advanced Voice Mode.” Unlike the older take-turns approach that waited for pauses, GPT-Live-1 can listen and speak at the same time (bidirectional). Users can interrupt it mid-sentence, and it responds more naturally and fluidly to the flow of conversation.
GPT-Live-1 is the default voice variant on the paid Go, Plus and Pro plans; the free tier uses the smaller GPT-Live-1 mini. It is available on iOS, Android and the web. For especially complex questions, GPT-Live-1 hands the task off to a more powerful text model in the background — initially GPT-5.5.
The model operates independently of the text tiers in the GPT-5.6 family (Sol, Terra, Luna) and serves a distinct purpose as a voice interface. Access is proprietary and runs exclusively through ChatGPT.
Example / practical relevance
In practice, GPT-Live-1 suits spoken, dialogue-based use: live translation, hands-free research, or conversations where the user spontaneously asks follow-ups or interrupts. The model tolerates background noise, can search the web while still speaking, and can display visual cards (for weather, stock prices, sports results).
For a marketing team, the voice channel is relevant for mobile research on the go or for accessibility — anywhere typing is impractical and a natural voice dialogue reaches the result faster.
Distinction from similar terms
GPT-Live-1 is a voice interface, not a text model — clearly distinct from the text models of the GPT-5.6 line (Sol as the strongest, Terra in the middle, Luna as the fastest and cheapest). The “Live” label denotes the real-time voice capability, not a particular text performance tier. The GPT-Live-1 mini variant is the pared-down, free edition of the same voice model.