Anthropic's Claude voice mode now uses any model and controls your apps

Anthropic is finally letting Claude’s voice mode stretch its legs. The company announced Thursday that users can now choose which AI model powers their spoken conversations—Opus, Sonnet, or Haiku—rather than being locked into the speedy but shallow Haiku. That’s a direct shot at the wall OpenAI’s own voice mode keeps hitting: it can sound wonderfully human but still can’t book a meeting.

Here’s where it gets genuinely useful. Claude’s voice mode will now tap directly into Gmail, Google Calendar, Slack, Canva, and Notion. You can ask it to draft an email, update a meeting slot, or spin up a doc, all without touching a keyboard. It picks the last text-chat model you used and defaults to its fastest version, which is a smart way to balance power and latency without forcing users to fiddle with settings.

The gap with OpenAI is becoming a gulf in terms of utility. While Sam Altman’s team focuses on making conversations feel more emotionally resonant, Anthropic is quietly stitching Claude into the fabric of your workday. The company also expanded multilingual support in beta earlier this year, now covering a handful of languages from Hindi to Brazilian Portuguese, though you’ll have to manually specify which one you’re speaking.

There is a catch. Anthropic didn’t touch the underlying voice model itself. So if you were hoping for more natural interruption handling or a less robotic cadence, you’re out of luck. Free users are also walled off, restricted to Haiku and a single app connection. Still, for anyone who finds OpenAI’s voice mode charming but useless, Claude just became the assistant that actually does something. That’s a bet on pragmatism over personality, and for now, it’s one only Anthropic is making.