OpenAI Rebuilds ChatGPT Voice to Listen While It Speaks
OpenAI shipped GPT-Live, a rebuilt ChatGPT voice architecture that keeps audio flowing continuously so deeper reasoning and tool calls no longer interrupt the conversation. Audio now moves through a dedicated fast path while reasoning runs asynchronously, cutting voice-session startup from six network round trips to one. OpenAI describes it as a turnless speech model built in six months, now rolling out in the desktop app and able to direct multiple Work or Codex agents by voice.
Why It Matters
Removing the turn-taking bottleneck is the difference between a voice assistant that pauses to think and one that behaves like a continuous collaborator — a meaningful step for hands-free, always-on agent control.