Anthropic's Claude Opus Finally Powers Voice Mode With Real Tool Access
Claude's voice mode now runs on Opus and Sonnet, can call your connected tools mid-conversation, and speaks 18+ languages on every plan
- Smarter models in voice: Claude voice mode now runs on Opus and Sonnet instead of Haiku, enabling real reasoning during spoken conversations.
- Tool access mid-conversation: Claude can now reach connected tools like email and calendar while you're talking, without switching to chat.
- 18+ languages on all plans: Multilingual support including Spanish, French, Hindi, and Japanese is now available to Free users too.
- Rolling out now: The update is live in public beta on mobile, desktop, and web today.
- Text-to-speech pipeline: Voice still runs on a conventional TTS stack (via ElevenLabs), not a speech-native model, so audio nuance is limited.
- Quota counts apply: Voice conversations on Opus or Sonnet consume your standard message quota; Free users get roughly 20-30 conversations per day.
Claude's voice mode just got a meaningful upgrade. Anthropic is rolling out a public beta that lets voice conversations run on its more capable models, Opus and Sonnet, reach tools already connected to Claude, and speak in 18 languages beyond English. The update is live today on mobile, desktop, and web.
The model upgrade that actually changes things
Until now, voice mode ran on Claude Haiku 4.5 regardless of what the model selector showed. That selector had sat inside the voice interface for roughly three weeks, but the choice was cosmetic: whichever option you picked, the session still ran on Haiku 4.5. That changes with this release.
Selecting Opus or Sonnet now routes voice conversations through those models rather than silently falling back to Haiku. Haiku is Claude's fastest, lightest model, optimized for speed and low cost. Opus and Sonnet sit above it in the reasoning hierarchy, handling longer chains of thought, multi-step tasks, and complex instructions significantly better.
In testing, interruption handling held up well. The model pauses the moment it hears you and resumes sensibly, tolerating long silences without cutting in early. Routing Opus or Sonnet into voice opens tasks that were awkward before: longer reasoning chains, tool-heavy requests handled entirely by speech.