Anthropic Expands Claude Voice Mode to Opus and Sonnet Models on All Platforms

Anthropic has expanded Claude's voice mode to run on its more powerful Opus and Sonnet models, moving beyond the Haiku-only setup that previously powered the feature. The update applies across all platforms where Claude operates: the mobile app, desktop app, and web chat interface.

Users can now switch between models mid-conversation, giving them direct control over the tradeoff between response speed and capability during voice interactions. The voice mode supports eleven languages and integrates with connected tools including Gmail, Google Calendar, and Slack, letting users compose and send emails through spoken commands.

Turn-Based vs. Full-Duplex

Claude's voice mode uses a turn-based system, waiting for the user to finish speaking before generating a response. This differs from competing approaches: OpenAI's GPT-Live uses full-duplex audio, allowing the model to speak and listen simultaneously, while Google's Gemini Live follows a similar full-duplex model but is currently restricted to smartphone use.

According to reporting from The Decoder, both OpenAI and Google's voice implementations feel more natural in practice due to more advanced speech processing pipelines. All three systems handle voice-based search queries competently, but Claude differentiates itself through tool integration — Anthropic is currently the only provider among the three that allows users to compose and save an email directly through voice interaction alone.

What This Means

This update signals Anthropic's bet that utility — specifically, deep integration with productivity tools like Gmail and Slack — can offset a technical disadvantage in raw voice interaction quality. Full-duplex audio, as used by OpenAI and Google, produces conversations that feel closer to natural human speech, since the model can interject, acknowledge, or adjust in real time rather than waiting for a complete pause.

By contrast, Claude's turn-based design is a more conservative architecture, likely easier to control and less prone to interruption errors, but it trades away some conversational fluidity. Extending Opus and Sonnet access to voice mode suggests Anthropic sees voice as a serious entry point for its most demanding users — those who want frontier-model reasoning applied to spoken tasks like drafting a nuanced email or interpreting calendar conflicts, not just quick lookups.

The real test will be whether tool integration outweighs interaction quality for everyday users. Voice interfaces succeed or fail on feel as much as function, and if GPT-Live and Gemini Live continue to feel more responsive, Claude's tool-calling advantage may only matter to a narrower set of power users who already rely on Anthropic's ecosystem for work tasks. For now, this is an incremental but meaningful upgrade rather than a shift in competitive positioning.