ChatGPT and Claude both upgrade voice
OpenAI brought full agent-control voice mode to the ChatGPT desktop app while Anthropic gave Claude voice mode model choice and app connections, the same week.
OpenAI rolled out its GPT-Live-powered voice mode to the ChatGPT desktop app globally on July 24, letting users direct multiple agents running in ChatGPT Work or Codex, and control their computer, with spoken commands, including real-time interruption handling and macOS screen-content awareness through a feature called Appshots. A day earlier, Anthropic updated Claude’s voice mode to let users pick between Opus, Sonnet, and Haiku instead of being locked to Haiku, and connected it to Gmail, Google Calendar, Slack, Notion, and Canva, so a spoken request can reschedule a meeting, draft an email, or create a document. Anthropic’s release is in beta on all platforms; it did not change the underlying voice model itself, so it still lacks some of the conversational handling, like interruption support, that OpenAI’s update added.
The two releases land the same week and split the same bet differently: OpenAI is positioning voice as an interface for controlling autonomous agents and your machine directly, while Anthropic is positioning it as a way to trigger the everyday apps you already use. Neither company frames voice as a chat novelty anymore, both are building it as a control surface for getting real tasks done hands-free.
This is part of a broader shift toward full-duplex, always-listening voice as the default AI interface rather than an add-on. If you build on either platform, the practical move is to test whether voice-triggered actions, especially agent control and third-party app connections, are reliable enough for your workflow yet, since both features are new enough that edge cases in interruption handling and multi-step command parsing are still likely.