OpenAI Unveils GPT-Live Voice Models for ChatGPT — Breakthrough
OpenAI on Wednesday rolled out GPT-Live, a new family of real‑time voice models that let ChatGPT listen and speak at the same time to create more natural, back‑and‑forth conversations. (investing.com)
The models use a full‑duplex architecture: they process incoming audio continuously while generating speech, so the agent can acknowledge a speaker with short interjections like “mhmm” or “yeah,” trade rapid exchanges, or hold silence when appropriate. That design replaces older turn‑based and cascaded pipelines that required users to finish speaking before the AI responded. (investing.com)
OpenAI is shipping two variants at launch — GPT‑Live‑1 and GPT‑Live‑1 mini — and is rolling them out to ChatGPT users worldwide on iOS, Android and the web. GPT‑Live‑1 will be the default for Go, Plus and Pro subscribers, while the mini version will serve free users; OpenAI says API access will follow soon. (investing.com)
Under the hood, GPT‑Live can offload complex searches, deeper reasoning, or multi‑step work to “frontier” models running in the background; OpenAI says it will use GPT‑5.5 for those background tasks at launch and will update the connected model as newer frontier models appear. Users can choose reasoning preferences — Instant for fast answers or Medium/High for more demanding thinking — and ChatGPT Voice can surface visual cards (weather, stocks, sports) while a conversation is ongoing. (investing.com)
In internal head‑to‑head evaluations, testers preferred GPT‑Live variants over the previous Advanced Voice Mode for five‑ to ten‑minute conversations, judging turn‑taking, interruptions, conversational flow and naturalness. More than 150 million people use ChatGPT Voice and Dictation features weekly, underscoring the scale of the update. (investing.com)
Separately, OpenAI has updated its Realtime API with newer models for developers building low‑latency voice agents: gpt‑realtime‑2.1 and a distilled gpt‑realtime‑2.1‑mini aimed at faster, lower‑cost realtime interactions. Those Realtime models reduce voice latency and add configurable reasoning and tool use for production voice agents. (developers.openai.com)
For builders, the combination of GPT‑Live in consumer ChatGPT and the Realtime API models means better barge‑in handling, shorter round‑trip delays and deeper integration between speech and reasoning — features that OpenAI says will unlock a new class of voice apps for customer service, assistants and interactive agents. (marktechpost.com)
OpenAI notes some limits at launch: GPT‑Live does not yet support voice paired with video or screen sharing inside ChatGPT, though the company is working to add those capabilities. Users who need the older voice features can still access previous ChatGPT Voice versions where available. (investing.com)
Original Source: https://www.investing.com/news/stock-market-news/openai-launches-gptlive-voice-models-for-chatgpt-93CH-4782124
Category:
Tags:
Publish Date: 2026-07-08 22:36:00