Anthropic just gave Claude’s voice mode a serious upgrade. The company is now letting you talk to its two most capable models, Opus and Sonnet, through voice, according to The Verge AI. Until now, voice mode ran only on Haiku, Anthropic’s fastest but least powerful model. That limit is gone, and the change turns voice from a quick-answer gadget into something you can actually work with.
The Verge AI reports that when Anthropic launched voice mode back in 2025, it was built for speed. Short questions, fast replies, minimal delay. But people didn’t use it that way. They started working “through real business problems,” and Haiku wasn’t designed for that kind of depth. As the company put it, that model “kept conversations quick, but not always deep.”
What’s new
Here’s what the upgrade actually delivers, based on The Verge AI’s reporting:
- Opus and Sonnet in voice. These are Anthropic’s deeper models, built for what the company calls “hard problem-solving.” They handle more complex responses and analysis, not just fast facts.
- Voice that takes action. The models can do things on your behalf. Turn a spoken conversation into a one-page pitch. Shift your calendar around if your train is running late. It’s voice that does work, not just talks.
- App integrations. Voice mode now reaches into Gmail, Slack, and Canva, so it can pull from and act inside tools you already use.
- Switch models mid-conversation. You can move between text and voice without breaking the thread, and swap models on the fly. Start with a quick Haiku chat, and if it sparks an idea, jump to Opus to go deeper.
- Nine new languages, out of beta. Non-English voice was beta-only until now. It’s fully available in French, German, Spanish, Hindi, Indonesian, Italian, Japanese, Korean, and Portuguese.
Why it matters
What stands out here is the shift in what voice mode is for. Most voice assistants still live in the ask-a-question, get-an-answer loop. Anthropic is betting people want something closer to a thinking partner they can speak to, one that can reason through a messy problem and then go execute on it. Putting Opus and Sonnet behind the microphone is what makes that possible. Haiku could keep up with the pace of speech, but it couldn’t carry the weight of a real work session.
The app integrations are the other half of the story. Voice that can only chat is a novelty. Voice that can rearrange your calendar, draft in Gmail, or build something in Canva starts to look like an actual assistant. The line between “talking to an AI” and “delegating to an AI” gets thinner.
The catch
A few things worth keeping in mind. The Verge AI’s report focuses on the capabilities, not the price, so it’s not spelled out which subscription tiers unlock Opus and Sonnet in voice. Anthropic typically gates its stronger models behind paid plans, so heavier voice use may cost more than the Haiku experience did. The deeper models also trade some speed for depth, which is the point, but worth expecting if you’re used to Haiku’s instant replies.
Anthropic also issued a correction on this one. An earlier version of the story said voice mode launched in 2026. It actually launched in 2025.
What comes next
This is Anthropic pushing voice from a feature into a workflow. The real test is whether people trust it to take action, not just answer. Booking, drafting, and rescheduling by voice only works if it works reliably, and that’s the bar Opus and Sonnet now have to clear. For the full breakdown, check the original report at The Verge AI.