Anthropic Upgrades Claude Voice Mode with App Integration and Model Picker
Artificial intelligence developer Anthropic has released a substantial upgrade to Claude‘s voice functionality, giving users the ability to connect voice commands directly to everyday workplace software and select higher-capacity language models for complex queries.
The update marks a distinct strategic departure from competitors like OpenAI. While recent voice updates to ChatGPT focused heavily on lower-latency conversational fluidity and human-like inflection, Anthropic‘s latest release prioritizes functional utility. Claude can now perform active tasks across integrated third-party platforms—including Google Calendar, Gmail, Slack, Notion, and Canva—allowing users to verbally instruct the assistant to edit calendar invites, draft emails, or generate new documents inside workflow software.
Under the new system, voice interaction is no longer restricted to Anthropic’s lightweight Haiku model. Users can now toggle between the full suite of Claude models—Opus, Sonnet, and Haiku—depending on the complexity of their task. By default, the voice mode automatically pairs with the fastest operational variant of whichever model the user last selected in their text chat session. This change addresses previous limitations where the speed-optimized Haiku engine struggled with multi-step reasoning, brainstorms for market research, or granular feedback on business pitches.
The company has also expanded global accessibility by rolling out multilingual capability in beta. The system currently supports ten languages: English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Brazilian Portuguese, and Latin American or European Spanish. However, automatic language detection is not yet active, requiring users to manually select their preferred language before starting a spoken conversation.
Despite these task-oriented upgrades, Anthropic did not alter the underlying audio generation stack for this release. Consequently, the voice mode lacks advanced conversational features found in competing tools, such as dynamic interruption handling or expressive speech variations. The feature is currently available in beta across all supported platforms, though free-tier accounts face restrictions, limiting usage to the Haiku model and a single connected application.
