OpenAI Introduces GPT-Live — A New Generation of Voice Models
OpenAI has unveiled GPT-Live, a new generation of voice models that support full-duplex communication and can handle resource-intensive tasks in the background.
OpenAI Introduces GPT-Live — A New Generation of Voice Models
Yesterday, OpenAI introduced GPT-Live — a new generation of voice models that now power the ChatGPT Voice mode.
Key Changes
The main change is the support for full-duplex. The model can listen and speak simultaneously, allowing it to naturally insert short affirmations like "uh-huh," maintain a fast-paced dialogue without pauses, or remain silent while you contemplate your response.
Additionally, GPT-Live separates the conversation itself from resource-intensive tasks. If a search or deeper reasoning is required, the model hands off that task to the flagship model in the background (currently GPT-5.5 at launch) and continues the conversation without interrupting the dialogue.
Available Versions
Two versions are available: GPT-Live-1 and GPT-Live-1 mini, which are already starting global deployment. GPT-Live-1 will be the default model for users on Go, Plus, and Pro plans, while GPT-Live-1 mini will be available for the free plan. Users will also be able to choose their reasoning level: Instant, Medium, or High. API support will come later.
GPT-Live is already available in European Union countries.
I won't lie, it looks really cool. 💃
Why it matters
AnalysisThis new model from OpenAI opens up new possibilities for interactive communication, making it more natural and efficient. The support for background tasks allows users to receive deeper responses without interrupting the dialogue.
Discuss in community
Share your questions and insights with developers