Anthropic upgrades Claude voice mode to Opus and Sonnet models, adds Gmail and Slack access

Claude voice mode gains Opus, Sonnet, plus Gmail and Slack

Anthropic just swapped the engine under Claude's voice mode. For the past year, talking to Claude out loud meant talking to Haiku, its smallest and fastest model. Haiku was built for speed, not depth: quick questions, quick answers, nothing that required sustained reasoning. As of today, voice mode runs on Opus and Sonnet, the same models people choose for complex written work. 1 2

The model swap matters. But the sharper move is what Claude can now reach while it talks. Voice mode gained access to connected apps including Gmail, Google Calendar, Slack, Canva, and Notion. Ask it mid-conversation to reshuffle your calendar, draft an email from what you just discussed, or drop a summary into a Notion document, and Claude can do it without breaking the flow of conversation. 2

That pairing is where Anthropic and OpenAI diverge. OpenAI's voice mode still cannot use external tools during a conversation, and that gap is the whole story. 2

The history explains the shift. When Anthropic launched voice mode in 2025, it ran exclusively on Haiku and was optimized for the shortest possible delay: ask a simple question, get a fast answer. 1 Users brought serious work to a voice assistant running on a model built for speed, not analysis. 1 Anthropic says the update was built for exactly those cases: longer conversations, feedback on communication style, working through a client pitch, brainstorming product research. 2 Spoken brainstorms can become structured pitch documents. A late train can trigger a calendar reshuffle, hands-free. 1 Users can also switch models mid-conversation and move freely between text and voice without starting over. Voice mode defaults to the fastest version of whichever model you last used in the text chat. 2

There are honest limits worth naming. Anthropic did not upgrade the voice model itself and has not disclosed what voice technology stack it uses. That means the improvements are in the thinking, not the talking. Conversational qualities like better interruption handling, which OpenAI prioritized in its recent update, may be unchanged here. 2

The new voice mode is available to all users in beta across platforms, but it comes with a tier wall. Free users are restricted to the Haiku model with only one connected app. The full Opus and Sonnet experience, along with broader app access, requires a paid plan. 2

Anthropic expanded language support as well. Voice mode now handles French, German, Spanish, Hindi, Indonesian, Italian, Japanese, Korean, and Portuguese alongside English, though users must manually select the language they want to speak. 1 2

This update landed weeks after OpenAI rolled out its own refreshed voice experience, and the two companies are working on different halves of the same problem. 2 OpenAI is pushing how natural a spoken AI conversation can feel. Anthropic is pushing how much of your digital life a spoken AI can act on while it talks. For anyone building or investing in this space, the signal is specific: the competitive frontier in voice AI has moved. It's no longer about who sounds best. It's about whose assistant can reason over your Slack threads, your inbox, and your calendar while you talk, and then do something about it. Whoever combines conversational fluency with that kind of live data access first will own a voice assistant that doesn't just answer questions. It gets things done.