Claude Voice Mode Finally Lets You Talk to Opus - and Your Apps

Claude Voice Mode just fixed the one complaint that made it hard to take seriously: it no longer runs only on Haiku. Anthropic's update lets you hold spoken conversations with any of its main models - Haiku, Sonnet, or Opus - so complex questions no longer get shallow, rushed answers.

If you pay for Claude and rely on voice while driving, cooking, or multitasking, this changes how useful the chatbot actually is day to day. Voice mode previously defaulted to Haiku, the smallest and fastest of Claude's four models, because lighter models keep conversations flowing naturally - but that speed came at the cost of fumbling harder queries.

Opus is the most capable model in the lineup, built for complex problem-solving, and having it behind your voice means spoken chats can now handle the same heavy lifting as typed ones. The only holdout is Fable, the top-tier model, which is not available in voice mode yet.

Switching works exactly the way it does in text chat, and the whole flow is designed to be flexible:

1. Open any conversation and pick Haiku, Sonnet, or Opus from the same model dropdown menu you already use in chat.
2. Tap the waveform icon to launch voice mode.
3. Mix input styles freely - start a thread by typing, switch to speaking halfway through, or even type while talking.

Your previous context carries over when you switch, so you never have to get Claude back up to speed. One thing to keep in mind: since speaking and typing run through the same conversation, time spent talking counts against your normal usage limits.

Beyond conversation, voice mode is powerful enough to make Claude act on your behalf with nothing but words. It works with connected apps including Gmail, Google Calendar, Google Docs, Slack, the digital organization tool Notion, and Canva.

In practice, that means you can tell Claude to push a meeting back when you are stuck in traffic, draft replies to whatever looks urgent in your inbox, or turn a rambling voice note into a one-page pitch. For busy professionals, that turns the chatbot into something closer to a hands-free executive assistant.

Spoken commands carry a higher risk of being misunderstood, and nobody wants an email firing off to the wrong person. Claude handles this by asking for confirmation before it executes any task on a connected tool, and voice mode inherits whatever access you already granted in typed chat - nothing more.

The update also makes Claude far more multilingual. Voice mode can now converse in French, German, Hindi, Indonesian, Italian, Japanese, Korean, Brazilian Portuguese, and Spanish for both Latin America and Spain.

You can switch languages mid-conversation simply by asking Claude to do so. If you want a permanent default, set it manually via Settings > General > Voice > Language.

Here is the catch: the full feature list sits behind the paid tiers. Free users stay locked to Haiku with just one connected app, though every new language is included at no cost.

Look at what Anthropic chose to give away versus what it gated, and the economics become obvious. Free users get all nine new languages - a feature that costs the company almost nothing extra per conversation - while Opus access stays paid, because serving its most capable model over real-time voice is genuinely expensive compute. The free tier is not a demo of the best experience; it is a funnel aimed squarely at the moment Haiku fails you mid-sentence.

The confirmation prompt before executing app actions is the quieter but more important signal. As voice assistants shift from answering questions to taking real actions in Gmail and Slack, a single misheard command becomes a real-world liability, and Anthropic is clearly designing for that trust gap before regulators or a viral mishap force the issue. Expect Fable voice support to arrive later as a paid-only differentiator - and if you are a free user who genuinely needs voice for work, the one-app, Haiku-only cap is engineered to make upgrading feel like your own idea.