OpenAI Launches GPT-Live-1 Voice Models for ChatGPT

OpenAI has introduced two new voice models — GPT-Live-1 and GPT-Live-1 mini — designed to make ChatGPT's spoken conversations feel more natural and less turn-based. The launch replaces Advanced Voice Mode and signals OpenAI's broader push to position voice as a core interface, not just a feature.

What's new

The headline change is full-duplex audio: the new models can speak and listen at the same time, rather than waiting for a user to finish talking before responding. GPT-Live-1 mini becomes the new default across ChatGPT, while paid-tier users get access to the larger GPT-Live-1 model. The voice mode also taps into GPT-5.5, giving it access to search, reasoning, and agentic capabilities during spoken conversations.

More than 150 million people currently use ChatGPT's Voice and Dictation features, giving OpenAI a large existing base to roll this out to.

The pitch: voice as a primary interface

Atty Eleti, OpenAI's ChatGPT Voice product lead, said he's had 30- to 40-minute conversations using the feature during walks, and argued that voice could become a primary interface to computing — particularly for managing complex, long-running agentic tasks rather than just quick queries.

Not without rough edges

A live demo of the translation feature in Hindi reportedly came across as unnatural and "bookish," with a heavy American accent — a reminder that voice quality outside English may still need work. OpenAI has not provided information on how widespread this issue is across other languages, so it's unclear whether this is an isolated demo hiccup or a broader pattern.

There have also been unconfirmed reports that OpenAI could launch AI-capable earbuds this year. OpenAI has given no official confirmation on any hardware plans, so founders and press alike should treat this as speculative for now.

A crowded conversational-assistant field

OpenAI isn't alone in chasing more natural voice interfaces. Apple and Amazon have both updated their assistants for better conversational flow and context handling. Sesame, founded by Oculus co-founder Brendan Iribe and Ankit Kumar, has launched assistants that hold more natural conversations while completing background tasks. Monogram, a startup focused on visual responses for more interactive assistants, raised $40 million in seed funding from DST and Lux Capital — a signal that investors see room for differentiation in this space beyond OpenAI's approach.

What's missing

OpenAI hasn't detailed pricing tiers or a specific rollout timeline distinguishing GPT-Live-1 from GPT-Live-1 mini. No technical benchmarks or head-to-head comparisons with competitor voice models have been shared either, making it hard to independently verify how much of a leap this represents.

Why founders should care

For founders building in AI, this launch is likely to matter in a few specific ways:

If full-duplex, always-on voice interaction becomes the norm, demand for natural, low-latency voice interfaces could increase — a trend startups building conversational tools may want to track closely.
OpenAI defaulting to voice suggests that agentic AI products may increasingly benefit from voice-first UX, particularly for long-running or multi-step tasks.
The rough Hindi translation demo hints that multilingual voice quality could remain an underserved market, potentially an opening for startups focused on non-English-speaking regions.
Renewed competitive activity from Monogram, Sesame, Apple, and Amazon suggests investor appetite for differentiated conversational assistants — whether through visual responses, task completion, or other angles — may be strengthening, though it's too early to say which approach will win out.