Claude voice mode stopped being a side feature on 23 July 2026. Until that day it ran on Haiku only, Anthropic’s fastest and simplest model, which meant talking to Claude got you a noticeably weaker Claude than typing to it did. That gap is now closed, and voice runs on Opus, Sonnet and Haiku across the mobile apps, the desktop apps and the web.
This guide covers what the feature actually does, which plan gets which model, and the exact steps to switch it on for the device in front of you. It also covers the part every other write-up misses, which is that Anthropic ships three separate things called “voice”, and they do not speak the same languages or bill the same way.
The Key Takeaways
- Voice mode is a beta feature on every plan, including Free, on iOS, Android, Mac, Windows and the web.
- Since 23 July 2026 it runs on Opus, Sonnet and Haiku instead of Haiku alone.
- Free accounts get voice, but pinned to Haiku and one connected tool.
- Voice mode covers 10 languages. It does not include Russian, Chinese or Ukrainian, although two sibling features do.
- Voice in the Claude app counts toward your usage limits. Voice dictation in Claude Code does not.
What Is Claude Voice Mode?
Claude voice mode lets you hold a spoken conversation with Claude instead of typing. You speak, Claude speaks back, and the transcript appears in the chat so you can carry on in text whenever you want. Anthropic still labels it a beta feature, and it is available on Free, Pro, Max, Team and Enterprise.
It runs on Claude Mobile for iOS and Android, on the Claude desktop apps, and on the web. Anthropic is blunt about the priority, though. Voice mode is “built to work best from your phone”.
The part that makes it more than a novelty is tool access. Voice mode reaches the same connected tools your text chats use, so you can ask it to check a calendar or draft an email without touching the keyboard. Anthropic’s own example is pushing a meeting on your Google Calendar back by 30 minutes.
What Changed in July 2026
Before 23 July 2026, voice mode was locked to Haiku no matter what you paid. You could be a Max subscriber running Opus in text, tap the voice icon, and quietly drop to the lightest model in the range. It worked for timers and quick questions and fell apart on anything that needed reasoning.
Anthropic’s announcement changed three things at once. Voice now follows your subscription, so paid plans get Opus and Sonnet as well as Haiku. Conversations start on whichever model you last used in text chat, using its fastest version, and you can switch models mid-conversation.
One model is still missing.
Claude Fable is not supported in voice mode, which matters if Fable is your default in text. Our guide to Claude Fable 5 covers what that model is built for.
Is Claude Voice Mode Free?
Yes, with two catches. Free accounts get voice mode, every language and the full interface, but they are pinned to Claude Haiku and can connect only one external tool. Paid plans add Opus and Sonnet in voice plus all of your connected tools.
There is a second restriction that catches people out. Free accounts cannot use voice mode inside Claude Cowork or Claude Code, although dictation still works in both. Here is how the tiers compare.
| Plan | Models in voice | Connected tools | Voice in Cowork and Code |
|---|---|---|---|
| Free | Haiku only | One | Não |
| Pro | Opus, Sonnet, Haiku | All connected | Yes |
| Max | Opus, Sonnet, Haiku | All connected | Yes |
| Team | Opus, Sonnet, Haiku | All connected | Yes |
| Enterprise | Opus, Sonnet, Haiku | All connected | Yes |
Voice conversations count toward your regular usage limits, so a long spoken session eats the same allowance a long typed one would. If you are weighing the tiers, our breakdown of whether Claude AI is free and our guide to Claude’s pricing tiers both go deeper on what each plan buys.
How to Turn On Claude Voice Mode
The control is the same sound wave icon on every platform. It just sits somewhere different on desktop than it does on your phone, which is why people assume the feature has not reached them yet.
On Mac Desktop and the Web
- Open the Claude desktop app or claude.ai and start or open a chat.
- Look at the lower right corner of the chat window and tap the sound wave icon.
- Allow microphone access if macOS asks, then start talking.
On iPhone and Android
- Open the Claude app and tap into any conversation.
- Tap the sound wave icon in the text input field, sitting next to the microphone icon, not in the corner of the screen.
- Pick a voice when Claude offers you the list.
- Speak normally, or use push-to-talk if you are somewhere noisy, which means holding a button while you talk and releasing it when you finish.
Changing the Voice and the Language
Claude offers several voice options, and you choose one the first time you start voice mode. To change it later, open your settings and scroll to the voice settings section, where you can also adjust the pace on mobile.
Language is a separate setting, and it does not follow your phone. Open Settings, then General, then Voice and Language. One caveat is worth knowing before you judge the results, which is that Anthropic still treats support for languages other than English as beta.
Which Languages Claude Voice Mode Supports
Anthropic’s July 2026 list covers 10 languages across 11 options, since Spanish appears twice for Latin America and Spain. The full set is English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Brazilian Portuguese and Spanish.
You will find figures of 9 and 18 elsewhere. Both count something else, and neither matches what Anthropic published.
The more useful comparison is against the other two Claude features that take spoken input, because their coverage is wider and the differences are where people actually get stuck.
| Language | Voice mode | Mobile dictation | Claude Code dictation |
|---|---|---|---|
| Czech | Não | Não | Yes |
| Danish | Não | Não | Yes |
| Dutch | Não | Não | Yes |
| English | Yes | Yes | Yes |
| French | Yes | Yes | Yes |
| German | Yes | Yes | Yes |
| Greek | Não | Não | Yes |
| Hindi | Yes | Yes | Yes |
| Indonesian | Yes | Não | Yes |
| Italian | Yes | Yes | Yes |
| Japanese | Yes | Yes | Yes |
| Korean | Yes | Yes | Yes |
| Norwegian | Não | Não | Yes |
| Polish | Não | Não | Yes |
| Portuguese | Yes, Brazilian | Yes | Yes |
| Russian | Não | Yes | Yes |
| Spanish | Yes, two variants | Yes | Yes |
| Swedish | Não | Não | Yes |
| Turkish | Não | Yes | Yes |
| Ukrainian | Não | Yes | Yes |
Does Claude Voice Mode Support Russian?
No, not in voice mode.
Anthropic’s July 2026 list leaves out Russian, and also Chinese, Ukrainian, Turkish, Polish and Dutch. That is the most common disappointment among people who switch voice on expecting parity with the text chat.
There are two workarounds and both are official. Dictation in the Claude mobile apps supports Russian, Turkish and Ukrainian in beta, so you can speak your prompt and read the answer. Claude Code supports all three plus Czech, Danish, Dutch, Greek, Norwegian, Polish and Swedish.
Claude Code Handles Voice Completely Differently
Anthropic calls the terminal feature voice dictation, not voice mode, even though the command line prints “Voice mode enabled” when you switch it on. The distinction is real. Claude Code transcribes your speech into the prompt box and never speaks back, so it is a typing replacement rather than a conversation.
The billing is the real surprise.
In the Claude app, voice eats your usage allowance like any other message. In Claude Code, transcription does not consume messages or tokens and does not count toward the limits shown in /usage, which makes it effectively free to dictate all day.
Transcription is also tuned for developers. Terms like regex, OAuth, JSON and localhost come through correctly, and your project name and current git branch are fed in as recognition hints automatically. If you work across devices, our guide to running Claude Code from your phone pairs well with this.
The Commands You Need
| Command | What it does |
|---|---|
/voice | Toggles dictation on or off and keeps your current mode |
/voice hold | Hold a key while you speak, release to finish. This is the default |
/voice tap | Tap once to start, tap again to send |
/voice off | Disables dictation |
Hold mode binds to the spacebar by default and has a short warmup, because Claude Code detects a held key by watching for repeat events from your terminal. Tap mode has no warmup but needs version 2.1.116 or later, so check with claude --version if the command is rejected. Recording stops on its own after 15 seconds of silence or two minutes total.
What Can Block It
Voice dictation needs a Claude.ai account. It will not run when Claude Code is authenticated with a raw API key, or through Amazon Bedrock, Google Cloud’s Agent Platform or Microsoft Foundry. Voice mode also refuses to start for organisations with HIPAA compliance enabled, which produces a policy message rather than an error.
It needs a local microphone too, so it does not work over SSH, in Claude Code on the web, or in VS Code Remote sessions including Dev Containers and Codespaces. The reason is simple enough, since your microphone is on your machine and the extension is running on the remote host.
Claude Voice Mode vs ChatGPT Voice
The honest headline is that ChatGPT is still ahead on the conversation itself. Anthropic changed which models power voice in July 2026 but did not change the voice model, so the turn-taking behaves exactly as it did before.
| Feature | Claude voice mode | ChatGPT GPT-Live |
|---|---|---|
| Conversation flow | Turn-based, you finish then Claude replies | Full-duplex, you can talk over it |
| Free plan | Yes, pinned to Haiku | Yes, on GPT-Live-1 mini |
| Paid models | Opus, Sonnet, Haiku | GPT-Live-1 |
| Session length | Not published | Up to 2 hours |
| Status | Beta | Default voice experience since 8 July 2026 |
Where Claude Wins
Claude’s advantage is what happens after you finish the sentence. Voice mode reaches Gmail, Google Calendar, Google Docs and Slack, so a spoken instruction can move a meeting or draft a reply rather than just producing text about one. Anthropic also names Canva among the connected tools in its announcement.
Where ChatGPT Wins
Interruption, and it is not close. Full-duplex means you can cut in mid-sentence and be understood, which is what makes a voice assistant feel like a conversation rather than a walkie-talkie. Our guides to ChatGPT Live and ChatGPT’s Advanced Voice Mode cover how that model got there.
When Voice Mode Does Not Show Up
Most failures are permissions, not rollout. On a Mac, open System Settings, then Privacy & Security, then Microphone, and switch on the app you are using. For Claude Code that means the terminal itself, not Claude.
If your terminal is not listed there at all, there is no toggle to flip and Anthropic documents a reset. Run tccutil reset Microphone with your terminal’s bundle identifier, quit the app with Command+Q rather than closing the window, then reopen it and trigger the prompt again.
Do not run that command without the identifier.
Without it, macOS revokes microphone access from every app on your Mac, including Zoom and Slack, and each one has to ask for permission again the next time you use it. Anthropic flags this in its own documentation, so it is a known trap rather than a rare edge case.
Two other checks are worth a minute. Garbled transcription, or transcription in the wrong language, usually means the dictation language is still set to English. On a work account, the block is an organisation policy rather than anything you can fix yourself.
A Short Timeline
Voice mode is younger than it feels. Anthropic started rolling it out in beta on the iOS and Android apps on 27 May 2025, with free accounts included from the start but capped at a couple of dozen conversations.
The terminal came much later. 3 March 2026 brought voice dictation to Claude Code, beginning with a small share of users and ramping over the following weeks.
Then came 23 July 2026, the update that matters most, which put Opus and Sonnet behind the microphone across mobile, desktop and web and ended fourteen months of voice meaning a weaker Claude.
Conclusion
Claude voice mode is worth switching on now in a way it was not a month ago. Putting Opus and Sonnet behind the microphone changes what you can reasonably ask it, and connected tools turn a spoken sentence into an action rather than a paragraph. It is still turn-based, so it will not feel like ChatGPT does.
Start on your phone, since that is what Anthropic optimised for, and set the language before you judge the transcription. If Russian, Chinese or most of Central and Eastern Europe is your working language, use dictation or Claude Code instead and check back after the next update.
Voice support now varies by app, by plan and by surface, which makes juggling assistants harder than it looks. Fello AI takes the other approach. It puts several leading models behind one native Mac app for $9.99 a month, so your choice of model is not decided by which vendor shipped a feature this month. Our comparison of Claude and ChatGPT and our roundup of the best AI apps for iPhone are the natural next reads.
Sources for this article are Anthropic’s voice mode announcement, the Claude help centre entry on voice mode, and the official Claude Code voice dictation documentation.
FAQ
Is Claude voice mode free?
Yes. Voice mode works on the Free plan across mobile, desktop and web. Free accounts are limited to Claude Haiku and to one connected tool, and they cannot use voice inside Claude Cowork or Claude Code. Paid plans add Opus and Sonnet plus all of your connected tools.
Does Claude voice mode work on Mac?
Yes. It works in the Claude desktop app and on the web, and you start it from the sound wave icon in the lower right corner of the chat window. Anthropic says the feature is built to work best from a phone, so the mobile experience is the more refined one.
Does Claude voice mode support Russian?
No. Anthropic’s July 2026 voice mode list covers English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Brazilian Portuguese and Spanish. Russian is supported in dictation on the Claude mobile apps and in Claude Code voice dictation, which covers 20 languages.
Does voice mode use up my Claude usage limits?
In the Claude apps, yes. Voice conversations count toward your regular usage limits in the same way typed messages do. Claude Code is the exception, because transcription there does not consume messages or tokens and does not count toward the limits shown in the usage command.
How do I turn on voice mode in Claude Code?
Run the slash command /voice in the terminal, then hold the spacebar while you speak. Add hold or tap to pick a mode, and note that tap mode needs Claude Code 2.1.116 or later. You also need a Claude.ai account, since it will not run on a raw API key.
