On Thursday, July 9, 2026, OpenAI introduced a significant advancement to the ChatGPT voice experience with the launch of the GPT-Live model. The new feature allows the system to listen to the user continuously, even while they speak, delivering more natural and direct interactions.

This announcement marks a crucial evolution, as GPT-Live was designed to operate in ChatGPT’s voice mode with the main new feature of listening and speaking at the same time. This functionality promises more fluid conversations without unnecessary pauses, a notable improvement over previous systems. From the same date, the GPT-Live-1 and GPT-Live-mini models are now accessible to users.

On the same topic: OpenAI expands voice control from GPT-Live to the desktop for coding

Full-duplex architecture improves the fluidity of conversations on ChatGPT

The development of GPT-Live is based on an innovative full-duplex architecture. Thanks to this configuration, the system can process the user’s audio uninterruptedly, even while speaking, and issue short responses when appropriate. An additional feature is the model’s ability to wait for the end of user intervention in case of pauses, before formulating a reply.

More on this story: AI is present in a third of websites created after the arrival of ChatGPT

This improvement allows ChatGPT to sustain the conversation flow continuously, even when faced with questions that require greater consideration or internet consultation. In these situations, GPT-Live preserves user interaction while directing the request to a more sophisticated model, significantly reducing wait times during voice calls and optimizing the experience.

Integration of visual elements enriches interaction with the model

GPT-Live’s capabilities are not restricted to spoken responses alone. The system will have the functionality to present visual cards with relevant information on topics such as weather forecasts, financial market quotes and sports results. This addition aims to convert the use of voice in ChatGPT into a more complete experience, enriched by visual data displayed directly on the screen.

Availability of new templates for ChatGPT users

OpenAI has established that the GPT-Live-1 model will be the default voice experience for subscribers of the ChatGPT Pro, Plus and Go versions. Users of the free platform will have access to the GPT-Live-mini model. This innovative voice mode will be implemented on both platforms: in the mobile applications and also in the web version.