OpenAI Launches Full-Duplex GPT-Live-1 Voice Models
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI released GPT-Live-1 and GPT-Live-1 mini, full-duplex voice models that listen and speak simultaneously, enabling natural interruptions and live translation without the speech-to-text/text-to-speech pipeline used previously.
- GPT-Live-1 mini becomes the default voice mode in ChatGPT for all users, while paid tiers gain access to the larger GPT-Live-1 model, which can route queries to GPT-5.5 for search, reasoning, and agentic tasks while continuing the conversation.
- ChatGPT Voice product lead Atty Eleti said he has had 30- to 40-minute-long conversations with the voice feature during walks, and told a briefing that OpenAI sees voice becoming 'the future interface to all kinds of work' including complex long-running agentic tasks.
- OpenAI said more than 150 million people currently use ChatGPT's Voice and Dictation features, and the new mode can present information visually and stay silent while absorbing context.
- Rivals Apple, Amazon, and startups like Sesame — founded by Oculus co-founder Brendan Iribe and Ankit Kumar — are also pursuing more conversational, expressive assistants with better context handling.
- OpenAI built in safeguards for age-appropriate teen responses and self-harm resources, but emphasized it is not positioning the new mode as an AI companion.
- The Hindi live translation demo showed a heavy American accent with unnatural, bookish phrasing, and OpenAI declined to specify which languages the mode is optimized for beyond 'most spoken languages.'
Why it matters: OpenAI is staking a claim to voice as the next primary computing interface by collapsing the traditional speech pipeline into one full-duplex model that retains access to GPT-5.5 reasoning — a technical leap that could differentiate ChatGPT from Apple, Amazon, and well-funded startups like Sesame and Monogram chasing the same conversational prize.
