Claude is no longer keeping its stronger reasoning models behind the keyboard. On July 23, 2026, Anthropic expanded Claude Voice Mode so paid users can speak with Claude Opus and Claude Sonnet, rather than being limited to Haiku. The update also lets voice conversations reach connected tools such as Gmail, Slack, Google Calendar and Canva, pushing Claude closer to an assistant that can discuss a task and help carry it out.
According to Anthropic’s July 23 product announcement, the upgrade grew out of how people were already using Voice Mode. Quick questions became longer business discussions, brainstorming sessions and planning conversations. Engadget’s report on the rollout adds an important detail: Claude still uses a turn-based voice system, meaning it listens, pauses, and responds rather than speaking and listening simultaneously.
What Changed In Claude Voice Mode?
Before this release, Voice Mode ran on Haiku because Anthropic wanted faster responses. That worked for lighter questions, but users began bringing tougher problems into spoken chats. Opus and Sonnet are now available for deeper conversations, and eligible paid users can change models without ending the chat.
The main additions are:
- Voice conversations can use Haiku, Sonnet or Opus.
- Users can switch models while the conversation is running.
- Claude can reach connected tools during a voice chat.
- More languages are supported, including Hindi, Japanese, Korean and Spanish.
- Users can move between voice and text without restarting.
Claude defaults to the model family last used in text chat and uses the fastest available version of that selection for voice. The Verge’s coverage of the Claude update notes that users can begin with a quick Haiku exchange and move to a heavier model when the discussion gets more demanding.
Can Claude Now Control More Apps Through Conversation?
Yes, but “control” needs context. Claude does not gain unrestricted access to every app on a phone or computer. It works through tools and connectors the user has linked, and available actions depend on what each connector supports.
Anthropic gives practical examples. A user running late can ask Claude to move a Google Calendar meeting by 30 minutes. A spoken client-pitch discussion can become a Canva one-pager. Claude can also review recent emails and prepare replies. The company says Claude asks for permission before using a connected tool.
The Google Workspace connector documentation shows the limits. Claude can search Gmail, read messages and create drafts, but it cannot send Gmail messages for the user. Google Calendar is more hands-on: Claude can create, update and delete events, manage attendees and respond to invitations.
Claude’s official X account also posted a video of the update, saying Voice Mode can now reach connected tools mid-conversation.
What Can Users Do, And Where Are The Limits?
The bigger story is not simply that Claude talks through stronger models. Voice is becoming another way to operate connected software. Anthropic already offers interactive connectors for services such as Asana, Canva, Figma, Slack and Box, while other connectors can retrieve information or take supported actions.
Still, this is not universal voice control. A connector must exist, the user has to authorize it, and Claude inherits that person’s existing permissions. For actions such as booking, buying or reserving through supported apps, Anthropic says users confirm the details before anything is finalized.
There are voice limits too. Engadget reports that Claude remains turn-based rather than fully duplex, so conversations may feel less fluid than systems built to listen and speak at once. Claude also does not automatically detect a language change. Users must tell it to switch or choose a language in settings.
Voice conversations count toward normal usage limits, which is worth remembering during longer sessions.
Who Gets The Upgrade And What Comes Next?
The updated Voice Mode is in beta on mobile, desktop, and the web, although Anthropic says it works best from a phone. Free users remain on Haiku and can use one connected tool. Paid users get expanded model choices and all of their connected tools.
That split shows where Anthropic is taking the product. Voice Mode is being positioned less like a hands-free question feature and more like a working interface for email, calendars, design tools, and team apps. The Indian Express also highlighted the productivity focus, particularly Gmail and Slack access during spoken conversations.
For now, the useful change is straightforward: a user can talk through a problem, decide what to do, and ask Claude to act without breaking the conversation. It remains permission-based and connector-dependent, but voice is moving closer to becoming a genuine work interface.
FAQs
Is Claude Voice Mode available to free users?
Yes. Free users get Voice Mode with Haiku, one connected tool, and all supported languages.
Can Claude send Gmail emails by voice?
No. Claude can search email and create Gmail drafts, but users must send them manually.
Can users change Claude models during a voice chat?
Yes. Paid users can switch among Haiku, Sonnet, and Opus through the model picker anytime.
Does Claude automatically detect spoken language changes?
No. Users must tell Claude to switch languages or select one in voice settings manually.
Does Claude ask before using connected tools?
Yes. Claude asks for permission before using connected tools during a voice conversation with you.
