OpenAI Launches GPT Live Voice Models for Natural Human-AI Interaction
TL;DR
OpenAI is reportedly rolling out GPT-Live-1 for ChatGPT. Go, Plus and Pro users get the full model, while free users receive a smaller Mini version. The main feature is full-duplex voice: the model can listen and speak at the same time, reducing awkward pauses, interruptions and rigid turn-taking. GPT-Live can hand off harder tasks to stronger models in the background, such as research, planning or complex reasoning, while keeping the conversation moving.
Nauti's Take
Treat this as a reported rollout until OpenAI confirms it in your account or official docs. Small teams testing voice workflows should first measure latency, interruptions, background noise handling, and handoff to stronger models in real support, sales, or coaching conversations.
Briefingshow
Voice becomes less like dictation and more like an interface. If ChatGPT can listen, respond and run tasks in parallel, AI moves closer to phone calls, meetings and assistant workflows. The hard test is reliability: noisy rooms, accents, emotion and high-stakes decisions can still expose gaps that a smoother voice experience may hide.
Tweets
Video
Sources