OpenAI Launches GPT-Live, a Voice Model for Real-Time Conversation

OpenAI has launched a new generation of voice models called GPT-Live, designed to make talking with ChatGPT feel like a real conversation between people. The two versions, GPT-Live-1 and the lighter GPT-Live-1 mini, are rolling out to users worldwide on iOS, Android and the web starting July 8, 2026.

The key change from ChatGPT's previous voice mode is the shift away from a turn-based model, where the assistant waited for the user to finish speaking. GPT-Live-1 can insert short acknowledgment cues like "mhm" or "right," jump into a quick back-and-forth exchange, or stay silent when the other person needs a moment to think.

How it works

The voice model itself handles the naturalness and flow of the conversation, but when a question requires web search, deeper reasoning, or more complex work, GPT-Live-1 hands the task off to the latest flagship model running in the background and returns to the conversation with a ready answer. This means users don't experience the typical pause while a query is being processed.

Also new is simultaneous translation, a feature that was impossible to implement in ChatGPT's previous turn-based system. Users can ask the assistant to translate their speech into another language on the fly, with minimal delay. The model can also generate visual responses, such as a weather forecast, a match score, or a stock chart, when the context of the conversation calls for it.

Market context

The GPT-Live launch coincides with an especially busy week for OpenAI, which also released its GPT-5.6 family of text models at the same time. Introducing a separate line of voice models shows that OpenAI treats voice interaction as its own competitive front rather than just an add-on to text models, as it goes up against rivals like Google and Anthropic, who are also building their own conversational tools.

The announced API access for GPT-Live opens the door for developers and companies building their own voice applications, from customer service to assistants on mobile devices, to use the model. That could fuel a new wave of products built around voice conversations with AI, including in markets outside the United States.

What it means for users in Poland

For Polish ChatGPT users, the key question will be how well the simultaneous translation feature handles the Polish language, and how quickly GPT-Live reaches local versions of the app. If real-time translation genuinely works smoothly, it could prove useful for business contacts or travel, though there is no confirmed data yet on translation quality for less widely spoken languages.

Companies planning to integrate voice AI features into their products should keep an eye on the announced API access for GPT-Live, which will give developers the same full-duplex conversation technology that OpenAI is now offering in ChatGPT.

Sources: OpenAI (openai.com), TechCrunch (techcrunch.com), MarkTechPost (marktechpost.com), Dataconomy (dataconomy.com)