OpenAI launches GPT‑Live voice models for ChatGPT

OpenAI rolled out two new conversational voice models—GPT‑Live‑1 and the smaller GPT‑Live‑1 mini—today, promising more natural, interrupt‑friendly interactions and replacing the existing Advanced Voice Mode in ChatGPT [2]. The upgrade matters because it ties voice to OpenAI’s latest text models, potentially expanding the voice interface to longer, more complex tasks for its growing user base.

Full‑duplex voice and longer conversations

The new models are “full‑duplex,” meaning they can speak while listening, allowing users to cut in mid‑sentence—a limitation of prior voice modes that often forced users to wait for a response before speaking [2]. OpenAI says this solves “interrupting users while they’re talking” and improves answer quality, while also letting the system stay silent until prompted, preserving context over extended exchanges. In internal testing, the voice feature supported 30‑ to 40‑minute conversations during walks, a stark increase from earlier brief interactions [2].

Integration with the latest text engine and market positioning

GPT‑Live‑1 routes queries to OpenAI’s newest text model, GPT‑5.5, for search, reasoning, or agentic tasks, ensuring voice responses benefit from the most advanced language capabilities [2]. By default, the mini version replaces the prior Advanced Voice Mode for all users, while paid subscribers gain access to the larger GPT‑Live‑1 model. OpenAI reports that more than 150 million people already use ChatGPT’s voice and dictation features, indicating a sizable audience ready for the upgrade [2].

Rivals are moving in parallel. Apple and Amazon have refreshed their assistants with better context handling, and startups such as Monogram and Sesame are adding visual and interactive elements to voice assistants [2]. OpenAI’s claim that voice could become the primary interface for complex, long‑running work underscores a strategic push to stay ahead in the emerging voice‑first computing space.

What to watch

• Hardware rollout – OpenAI hinted at AI‑enabled earbuds later this year but gave no details.
• Model performance benchmarks – Future releases may include latency or accuracy metrics compared to Apple Siri or Amazon Alexa.
• Paid tier adoption – Uptake of GPT‑Live‑1 among ChatGPT Plus users will signal commercial traction.

The launch positions OpenAI’s voice as a more conversational, context‑aware layer on top of its strongest language models, but the real test will be whether developers and users adopt the full‑duplex capability at scale and how quickly competitors respond with comparable features.