Users can switch models mid-conversation and toggle between voice and text without starting a new chat, increasing flexibility during discussions.

The rollout is currently behind a feature flag and not publicly available yet; a formal post will likely follow once the functional switch is public.

The rollout is in beta, with ongoing evaluation of reliability and secure connectivity to user data as users move beyond beta.

The update introduces tool integration with Claude’s voice, contrasting with OpenAI’s voice approach which centers more on conversational style and does not yet offer broad tool usage.

The Claude blog is the primary source for official details and related feature references.

Accessibility is in beta for all users, with free-tier users limited to Haiku and a single connected app, while paying subscribers gain full access to Opus and Sonnet.

Voice mode remains turn-based rather than fully duplex, processing speech and response in sequence.

Opus and Sonnet are designed for deeper reasoning and longer, more analytical outputs, including tasks like creating a one-page pitch or rescheduling events.

Anthropic has updated Claude’s voice mode to support more capable models—Opus, Sonnet, and Haiku—with the last-used model defaulting to its fastest version for ongoing conversations, enabling longer, more capable interactions.

The update enhances voice mode for longer discussions by offering features like feedback on communication style, aiding with client pitches, and brainstorming for product market research.

Claude’s voice mode can access external apps (Gmail, Google Calendar, Slack, Canva, Notion) to perform actions like updating meetings or drafting emails directly from voice queries.

Choosing heavier models trades speed for deeper reasoning, with latency considerations still important for voice conversations.