OpenAI has launched a new generation of voice models called GPT-Live, designed to make conversations with ChatGPT feel more natural by allowing the AI assistant to listen, speak and process information simultaneously.

The company introduced GPT-Live-1 and GPT-Live-1 mini as replacements for ChatGPT’s older advanced voice mode. The models are designed to reduce interruptions, respond more naturally during conversations and handle tasks such as research, reasoning and translation in the background without forcing users to wait for the AI to finish processing information.

GPT-Live Can Listen And Speak At The Same Time

At the core of GPT-Live is a full-duplex architecture that allows the AI model to listen and speak simultaneously.

Unlike older voice systems that operated in turns and waited for users to stop speaking before responding, GPT-Live continuously processes information during a conversation.

According to OpenAI, the technology allows the model to decide several times every second whether it should speak, continue listening, pause, interrupt or use a tool.

The model can also respond with conversational acknowledgements such as “mhmm”, “yeah” or “got it” while a user is speaking, mimicking the type of feedback commonly used during human conversations.

OpenAI CEO Sam Altman said the improved experience could change his own preference for typing instead of speaking to AI assistants.

“GPT-live (next-generation voice) launches today in ChatGPT. It feels magical and ‘real’. I have always preferred typing to talking to an AI, now I think that’s going to shift,” Altman wrote on X.

GPT-5.5 Handles Complex Tasks In The Background

GPT-Live separates the voice conversation from tasks that require deeper reasoning, searches or more complex agentic capabilities.

When such a task is required, the voice model can hand it over to another system operating in the background, which at launch is GPT-5.5. Once the task is completed, the results can be brought back into the ongoing conversation.

During a demonstration before the launch, GPT-Live simultaneously translated speech into Hindi while the presenter continued speaking before joining the conversation with the completed translation.

The upgraded voice mode can also display visual cards for information such as weather, stocks and sports while maintaining support for search, memory, images and file uploads.

GPT-Live Models Outperform Older Voice Mode

OpenAI said GPT-Live-1 and GPT-Live-1 mini performed better than ChatGPT’s older advanced voice mode in human evaluations and technical benchmarks.

In matched conversations lasting between five and 10 minutes, GPT-Live-1 recorded a preferred response rate of 75.7 per cent compared with advanced voice mode.

GPT-Live-1 mini achieved a preferred response rate of 69.2 per cent in the same evaluations.

According to The Verge, OpenAI research lead Kundan Kumar described GPT-Live-1 as the company’s “smartest voice model” to date.

More than 150 million people use ChatGPT’s Voice and Dictation features every week for activities including everyday assistance, language practice, bedtime stories and general conversations.

OpenAI Expands Safety Testing For GPT-Live

OpenAI said it had expanded safety testing for the GPT-Live models, using audio-native and synthetic evaluations across areas including self-harm, psychosis and mania, emotional reliance, violence and sexual content.

The company said safeguards built into the models can redirect conversations towards safer responses, display safety messaging or end voice conversations in higher-risk situations.

OpenAI has also introduced protections for teenage users, including age-appropriate behaviour and parental controls.

The company said GPT-Live is designed for conversation rather than voice impersonation. The system uses predefined voices and includes safeguards aimed at preventing the imitation of real people’s voices.

The launch comes as OpenAI faces lawsuits alleging that ChatGPT harmed users.

GPT-Live Rolling Out To ChatGPT Users

OpenAI has confirmed that GPT-Live is being rolled out to ChatGPT users on iOS, Android and the web.

GPT-Live-1 will be the default voice model for ChatGPT Go, Plus and Pro subscribers, while GPT-Live-1 mini will be available by default to free users.

OpenAI also plans to make GPT-Live available through its API, with developers and enterprises able to register for notifications about access.

The company said the models have been optimised for some of the most widely used languages on ChatGPT, although users may still encounter non-native accents or gaps in fluency in certain languages.

At launch, GPT-Live does not support voice conversations alongside video or screen sharing within ChatGPT. Those capabilities will continue to be available through the platform’s older voice modes.

Long or Short, get news the way you like. No ads. No redirections. Download Newspin and Stay Alert, The CSR Journal Mobile app, for fast, crisp, clean updates!

App Store –  https://apps.apple.com/in/app/newspin/id6746449540 

Google Play Store – https://play.google.com/store/apps/details?id=com.inventifweb.newspin&pcampaignid=web_share