Anthropic Upgrades Claude Voice Mode With Opus and Sonnet
by Pritam Singh - 14 hours ago - 4 min read
Anthropic has upgraded Claude’s voice mode to work with its Opus and Sonnet model families, removing one of the main limitations of the feature since its launch.
Claude’s spoken interface previously relied on Haiku, Anthropic’s fastest but less capable model. The update means paid users can now use more advanced models for longer discussions, problem-solving and tasks involving connected workplace applications. The upgraded voice mode became available on July 23, 2026, across Claude’s mobile, desktop and web apps.
The update at a glance
| Feature | Previous experience | Updated experience |
|---|---|---|
| Available model families | Haiku | Haiku, Sonnet and Opus |
| Model switching | Limited | Available during a voice conversation |
| Connected applications | Basic connector access | Can retrieve information and perform approved actions |
| Platforms | Claude apps | Mobile, desktop and web |
| Voice-language options | Mainly English with limited beta access | 11 listed language and regional options |
| Free-plan access | Haiku | Haiku, one connected tool and all supported languages |
| Paid-plan access | Plan-dependent | Expanded models and connected tools |
The feature remains in beta, and voice conversations count against the normal usage limits included with each Claude subscription.
Voice mode now follows the model used in text chat
When a user starts voice mode, Claude initially selects the same model family most recently used in text chat. It then automatically uses the latest and fastest available generation of that model to keep spoken responses relatively smooth.
Users can also open the model selector during a conversation and move between Haiku, Sonnet and Opus without creating a new chat. The existing conversation context is preserved when switching between text and voice. Claude Fable, however, is not currently supported in voice mode.
This changes the type of work Anthropic expects people to complete through speech. Haiku remains suitable for quick questions, while Sonnet and Opus are positioned for more demanding discussions such as reviewing a business proposal, examining competing ideas or preparing for an important meeting.
Anthropic said users had already started applying the original voice feature to longer business conversations, even though Haiku had primarily been selected to prioritise response speed.
Claude can move from conversation to action
The update is not limited to generating more detailed spoken answers. Voice mode can access tools connected to a Claude account, including Gmail, Google Calendar, Google Docs and Slack.
A user could ask Claude to summarise recent emails, draft replies or check a calendar before changing a meeting time. Anthropic also demonstrated workflows involving Canva, such as turning a spoken discussion about a client pitch into a one-page document. Claude requests permission before using a connected tool to take an action.
This gives Claude’s voice interface a more productivity-focused position. Instead of operating only as a spoken chatbot, it can combine a conversation with information stored in other services and carry out approved tasks.
There are still plan-based restrictions. Free users receive access to Haiku and one connected tool, while paid plans can use additional eligible models and more connected services.
Eleven language and regional options are available
Anthropic has also expanded the languages supported by voice mode. The current list contains 11 selectable language and regional options: English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Brazilian Portuguese, Latin American Spanish and Spanish from Spain.
Users can change languages through voice settings or ask Claude to switch during a conversation. However, Claude does not automatically identify the language being spoken. Users must manually select it or request the change, and the language previously chosen for Claude’s interface does not automatically transfer to voice mode.
Anthropic’s help documentation still describes non-English voice support as being in beta.
The upgrade improves reasoning, not the underlying voice system
The most important distinction is that Anthropic has upgraded the language model processing the conversation rather than announcing a completely redesigned speech engine.
The company has not disclosed the technical voice stack behind the feature, and this release does not claim major changes to speech generation, interruption handling or the naturalness of Claude’s voice. The practical improvement should instead come from allowing spoken requests to reach models capable of deeper reasoning and more complex tool use.
That makes this a capability update rather than a voice-quality overhaul. Users may receive stronger analysis and better handling of complex requests, but they should not assume the update automatically delivers lower latency or more human-like speech.
For Anthropic, the change moves Claude voice mode closer to the intelligence already available in its text interface. It also turns voice into a more useful entry point for work that begins as a conversation but ends with an email, calendar change, research task or finished document.