OpenAI shipped ChatGPT Voice to its macOS and Windows desktop apps on Thursday evening, rolling it out globally as a full release rather than a preview. It runs on GPT-Live, the full-duplex voice model OpenAI launched on mobile on July 8, which listens and speaks at the same time instead of taking rigid turns.
The actual feature
Voice is being positioned as a supervisor, not a dictation tool. It can start new tasks, check the status of work already running, ask questions about agents, and adjust agents executing in parallel threads — all while the user is doing something else. On-screen actions route through OpenAI's existing Computer Use tool, operating within whatever permissions Work or Codex already hold. It reaches local files and connectors including Slack, GitHub and Notion.
One app, three surfaces
ChatGPT Work launched on July 9, when OpenAI folded Codex into the ChatGPT desktop app. Chat, Work and Codex are now three surfaces of one program. Work takes an outcome, decomposes it, runs for hours across connected apps and returns finished documents, spreadsheets, slides or web apps rather than chat replies. Fortune reports it reached roughly 10 million users after launch.
Where and how
Available on Plus, Pro, Business, Edu and Enterprise — not Free, and not on the web. It can also be driven from the ChatGPT iOS app through the Remote pairing feature; Android is "soon." A user-programmable global hotkey launches it. On macOS, an opt-in Screen context setting enables "Appshots," letting ChatGPT reference the frontmost window.
The competitive read
The merged desktop app is aimed at Claude Desktop, and the background-task behavior mirrors Anthropic's Claude Cowork; Codex competes with Claude Code, which added push-to-talk voice in March. OpenAI published no voice-specific new safeguards and no cap on parallel agents.