Claude Voice Mode Now Runs on Anthropic's Most Capable Models

Anthropic announced on Thursday that its voice mode now runs on the company's Opus and Sonnet models instead of only the Haiku model that powered the feature since 2025. The company said voice mode now starts with whatever model a user last selected in text chat so a spoken session inherits that model's intelligence automatically. Users can also switch between Haiku, Sonnet, and Opus mid-conversation using a model picker. The update also lets voice mode reach into connected apps such as Gmail, Google Calendar, Slack, Canva, and Notion to turn spoken commands into actions. Free users remain restricted to the Haiku model and a single connected app while Sonnet and Opus are reserved for paying subscribers. Multilingual input is included across ten languages though users must name the language before speaking it. The model picker does not include Claude Fable 5 which stays out of voice for now. Anthropic declined to detail the underlying voice stack it uses. The upgrade closes a gap with OpenAI's ChatGPT and Google's Gemini which offered advanced voice features earlier.