OpenAI launched ChatGPT Voice on the desktop app on July 23, 2026, letting users control their computer and direct multiple AI agents running inside ChatGPT Work and Codex simultaneously using only their voice. The feature, powered by GPT-Live’s full-duplex architecture, is rolling out globally on macOS and Windows to Plus, Pro, Business, Edu, and Enterprise plan subscribers.
“ChatGPT Voice is now in the desktop app. Control your computer and direct multiple agents running in ChatGPT Work or Codex, using just your voice. It’s powered by GPT-Live, so it can speak, listen, and coordinate work in the app at the same time. Rolling out globally today on macOS and Windows to Plus, Pro, Business, Edu, and Enterprise plans.” — OpenAI, official X account, July 23, 2026
What Is GPT-Live and Why It Matters for Agent Control
GPT-Live is OpenAI’s full-duplex voice model — it speaks and listens simultaneously rather than taking turns in a call-and-response pattern. Until this update, GPT-Live powered voice interactions only in the ChatGPT mobile app. The desktop launch extends it to the workspace environment where most business users run their AI workflows.
ChatGPT Work is OpenAI’s multi-agent orchestration tier, designed to run multiple agents in parallel on business tasks. Codex is OpenAI’s agentic coding tool. Both can now receive voice instructions through the same session without switching between keyboard inputs or conversation threads.
What the Desktop App Does That Mobile Couldn’t
On macOS, ChatGPT Voice incorporates a feature OpenAI calls “Appshots” — the AI analyzes the frontmost window alongside local files, codebase structures, and active plugins, giving agents screen context alongside verbal instructions. This capability is specific to the macOS desktop build and is not available in the mobile app.
Codex voice control also works from the iOS app via a paired remote connection, with Android support confirmed as coming soon, according to 9to5Mac’s coverage of the rollout.
Why This Changes Multi-Agent Workflows for Business Users
Before this update, directing multiple AI agents required typing into separate conversation threads — one command per agent, context switching between windows. ChatGPT Voice on desktop lets a user verbally check in on agent progress, redirect tasks, and start new work while agents run in the background, in the way a manager would direct a team of people rather than writing individual memos.
VentureBeat characterized the launch as “agentic coding goes hands-free,” noting the GPT-Live full-duplex architecture as the technical basis for simultaneous listen-and-coordinate behavior.
OpenAI’s move arrived on the same day Anthropic updated Claude’s voice mode — July 23, 2026 — to use Sonnet and Opus models instead of Haiku-only, indicating both companies are prioritizing voice-driven AI productivity for business users on the same competitive timeline.
For business users evaluating best AI agents for business, voice-native agent coordination on the desktop represents a significant workflow shift — one where verbal instruction replaces manual switching between agent threads. For a broader view of tools in this space, see AI tools for business.
For Context: OpenAI’s Voice and Agent History on witho2.com
- GPT-Live voice model — our previous coverage of GPT-Live’s full-duplex architecture and initial mobile launch
- ChatGPT Work — how OpenAI’s multi-agent orchestration tier works for business
Our Take
OpenAI is betting the bottleneck for business AI adoption is not capability — it is the effort of managing multiple agents by keyboard. Voice control on the desktop is the fastest bridge to the “AI employees” pitch. Anthropic’s same-day voice upgrade confirms both companies reached the same conclusion simultaneously: the race for enterprise AI adoption runs through voice-first interfaces in the workspace, not the mobile app.
