Product Launch2 min read

Claude Voice Mode Now Works With Smarter Models

July 23, 2026Synthesized from 1 source: The Verge

Anthropic has upgraded Claude's voice mode to run on its more capable Sonnet and Opus models, added connections to Gmail, Slack, and Canva, and expanded to ten languages, closing a gap that made voice useful only for quick, surface-level questions.

Claude's voice mode launched in May 2025, and it had a deliberate constraint baked in from the start. To keep responses fast, Anthropic routed all voice conversations through Haiku, the smallest model in the Claude family. That made it fine for simple questions, but thin for anything more demanding.

Anthropically has now changed that. Sonnet and Opus, the models Claude users typically rely on for serious writing, analysis, and complex reasoning, are now available in voice mode. You can pick your model before a conversation, or switch during one. If a quick exchange with Haiku sparks a bigger question, you can move to Opus right there without starting over.

The app connections matter here. Voice mode can now pull context from Gmail, Slack, and Canva, provided you connect those accounts. Paid subscribers on the Google Workspace plan also get access to Google Calendar. In practical terms: you could speak a request, have Claude check your inbox for the relevant thread, and come back with a draft or a decision, hands-free. That is a different category of usefulness than asking what the capital of France is.

Language support has also expanded. Voice mode now works properly in French, German, Spanish, Hindi, Indonesian, Italian, Japanese, Korean, and Portuguese, after those languages spent time in limited testing.

The honest comparison to ChatGPT is worth making. ChatGPT's voice system can listen and respond simultaneously, so it handles interruptions and overlapping speech the way a phone call does. Claude still operates in turns: it listens, stops, thinks, and then speaks. That turn-based structure makes conversations feel less fluid. Claude also cannot detect a mid-sentence language switch on its own. These are real gaps, and they are not fixed by this update.

What this update does fix is depth. Before today, using Claude's voice mode for a real business conversation was like asking your most knowledgeable colleague to only ever give you one-line answers. Now the full capability is available in spoken form.

For professionals who already use Claude for analysis or drafting, this is worth testing. The practical move is simple: try it for one actual work task this week, something you would normally do by typing, and see whether speaking it is faster and whether the output holds up. If it does, you have found a faster input method. If the turn-based pacing feels awkward, you will know voice is not yet your preferred interface for Claude. Either answer is useful.

Stay informed

Get AI intelligence like this delivered to your inbox.


You May Also Find Valuable