Anthropic Expands Claude Voice Mode to Opus, Sonnet
24 Jul 2026
Claude's voice mode grows up
Anthropic has expanded Claude's voice mode beyond its original Haiku-only setup, giving users the option to run voice conversations on Opus and Sonnet models as well. The update, rolled out on 2026-07-23, also brings new app integrations, letting Claude's voice mode connect with Gmail, Google Calendar, Slack, Canva, and Notion.
Until now, voice mode was restricted to Claude Haiku, Anthropic's lightest model. The upgrade means users can now access more capable models for voice-driven tasks, and can switch models mid-conversation without restarting. Users can also toggle between text and voice mode on the fly.
What's new, concretely
- Model choice: Opus, Sonnet, and Haiku are all now available in voice mode, up from Haiku-only previously.
- App integrations: Gmail, Calendar, Slack, Canva, and Notion connections let voice mode take actions across a user's workflow tools.
- Multilingual support: Added in beta earlier this year, though sources differ on the exact count — TechCrunch lists 10 supported languages (including English), while The Verge lists 9 (not counting English).
- Free tier limits: Free users are capped at the Haiku model and a single connected app.
- Availability: The new voice mode is in beta across platforms, open to all users.
How this stacks up against OpenAI
The timing is notable: OpenAI recently rolled out a new family of conversational models along with an updated ChatGPT voice mode. According to the report, OpenAI's voice mode cannot yet use tools to complete tasks — a gap that Claude's updated voice mode, with its app integrations, appears designed to exploit. This tool-use capability may be Anthropic's clearest differentiation point in the voice AI race right now.
One thing Anthropic has not changed is the underlying voice model itself. The company hasn't detailed what voice stack powers Claude's voice mode, and no changes were made to it in this release — a gap that leaves open questions about voice quality, latency, or reliability compared to competitors.
Why founders should care
For founders building on top of conversational AI, this update likely signals a few shifts worth tracking:
- Anthropic's push into Opus and Sonnet for voice suggests the company sees rising demand for more capable, professional-grade voice interactions — not just casual chat.
- The new app integrations (Gmail, Calendar, Slack, Canva, Notion) could indicate an opportunity for startups to build voice-driven productivity tools on top of Claude, especially if Anthropic continues expanding its integration list.
- The tool-use advantage over OpenAI's voice mode may matter for founders choosing a platform for agentic, task-oriented voice products — Claude's approach appears more geared toward getting work done, not just conversing.
- Because this is a beta rollout across all users, founders should expect possible instability or changes before a broader stable release, and should build with that in mind.
What's still unclear
Several gaps remain in the public record. There's no detail on pricing for Pro, Team, or Enterprise tiers beyond the free-tier restrictions, no usage or adoption metrics since voice mode's 2025 launch, and no side-by-side comparison of voice quality or latency against OpenAI's updated offering. The multilingual language count also differs between sources, so founders evaluating international deployment should treat the exact language list as unconfirmed pending Anthropic's own documentation.
The bottom line
Anthropic is betting that voice mode's future lies in flexibility — swappable models, mid-conversation switching, and real app integrations — rather than just better speech synthesis. If this bet pays off, it could open a lane for founders building voice-first productivity and workflow tools on Claude, particularly where task execution matters more than pure conversational polish. But with the voice stack undisclosed and the rollout still in beta, it's a space to watch closely rather than commit to irreversibly just yet.