OpenAI launches GPT-Live: ChatGPT can now listen and talk at the same time

On July 8, 2026, OpenAI launched GPT-Live, a new generation of voice models for the ChatGPT platform. The main change is full-duplex conversation: the model can listen and speak at the same time, without requiring pauses between the user's turns.

How it works

In the previous version of its voice interface, ChatGPT used a half-duplex mode: the user speaks, the model waits for a pause in speech, then responds. This felt unnatural. GPT-Live removes that limitation: the model processes the incoming audio stream in real time and can respond without waiting for the user to finish the sentence.

This brings the interaction closer to natural human conversation, where people often start replying before the other person has completely finished speaking.

Two model versions

OpenAI released two versions:

• GPT-Live-1: the full-featured model with all of ChatGPT's capabilities
• GPT-Live-1 mini: a lighter version for mobile devices and devices with limited resources

Both versions have been available to all ChatGPT users worldwide since launch (July 8, 2026).

What this means for conversational AI

Full-duplex conversation is a significant step toward natural interaction with AI. Until now, most voice assistants (Siri, Alexa, Google Assistant) have worked in half-duplex mode. OpenAI shows that full duplex is technically feasible on large language models.

It also shows that companies are investing in better conversational interfaces, probably because voice is becoming the main way of interacting with AI on mobile devices.

Why it matters

A mobile AI assistant that talks like a human, without awkward pauses, could attract millions of users. This is especially relevant in markets where a text interface is less convenient (developing countries, older users, multitasking).

CPU/GPU requirements for full-duplex processing are higher than for half-duplex, but the lighter GPT-Live-1 mini makes it possible to deploy the model on mobile devices.

What it means

Full-duplex conversation in ChatGPT is an intermediate step toward personal AI assistants that work like natural conversation partners. The industry is moving toward reducing friction between humans and models. OpenAI shows that the line between a "talking machine" and a "smart assistant" is fading.