Skip to content
NewsProduct launch

Anthropic opens Claude voice mode to Opus and Sonnet, adds Gmail, Slack, and Calendar connectors

· by Pondero Newsdesk

The short version

Anthropic expanded Claude voice mode on July 23 to run on Opus and Sonnet alongside Haiku, added live connectors for Gmail, Google Calendar, Slack, Canva, and Notion, and extended language support to 11 languages.

Anthropic opens Claude voice mode to Opus and Sonnet, adds Gmail, Slack, and Calendar connectors

Anthropic's Claude voice mode ran on Haiku alone since launch, fast but capped at the company's least capable model tier. On July 23, 2026, Opus and Sonnet joined the interface, five productivity apps gained live connectors, and voice support expanded to 11 languages across all plans, per Anthropic's product blog. For paid subscribers, a brainstorm session can now end with a calendar rescheduled, a Notion doc drafted, and a Slack message sent, all without leaving voice.

What changed

Model access

Opus and Sonnet are now available inside Claude voice sessions alongside Haiku. Switching between them requires only a tap on the model picker; sessions do not restart. Voice mode defaults to whatever model the user last selected in text chat, so moving between voice and text preserves continuity. Per Anthropic, the interface runs the fastest variant of whichever model is chosen so that turn-taking stays smooth. Free-plan users are locked to Haiku; paid subscribers can access all three tiers.

App connectors

Five services gained live connectors in this release: Gmail, Google Calendar, Slack, Canva, and Notion. A user running late to a meeting can ask Claude to push it by 30 minutes directly on Google Calendar. A conversation about a client pitch can produce a Canva one-pager. Claude requests permission before touching any connected service, and connections are configured in Settings under Connectors across the mobile, desktop, and web apps. Free users can activate one connected tool; paid subscribers get all five.

Language coverage

Voice mode now supports 11 languages: English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Portuguese (Brazilian), Spanish (Latin America), and Spanish (Spain). All plan levels get full language access. Switching requires an explicit verbal request or a selection from the voice settings language picker. Claude does not auto-detect spoken language, and voice-mode language settings are separate from text-chat settings and do not carry over between modes.

The feature launched in beta to all Claude chat users on mobile, desktop, and web, though Anthropic noted it works best from a phone. Voice conversations draw from the same usage limits as text sessions.

Why it matters

The Haiku constraint was the binding limit on session depth. Haiku is built for low latency, not extended reasoning. Users who tried working through a real problem in voice got fast responses but not the analytical follow-through that makes Claude useful in long text sessions. Opus and Sonnet bring multi-step synthesis and follow-up questioning into the voice interface. That gap is closed for paid subscribers.

For operators who run Claude on paid plans, voice mode becomes a viable input channel for work that previously required a keyboard. Running a decision review, stress-testing a pitch argument, or thinking through a hiring call out loud was already possible in voice, but the output quality was capped. Lifting that cap matters most for sessions that run long or involve competing considerations rather than quick factual lookups.

The connector roster carries a separate signal. Anthropic chose five tools that sit at the center of most knowledge workers' days (calendar, inbox, messaging, document creation) rather than building broad shallow integrations across dozens of services. The design, where Claude asks permission before each action and operates inside an ongoing conversation rather than as a background automation trigger, is reactive rather than pre-defined. That is a different architecture from workflow tools like Make or Zapier, which execute sequences set up in advance. Claude's connector model responds to the conversation as it unfolds.

Context and reactions

Anthropic said in the announcement that users had started holding longer, problem-focused sessions after hands-free conversation launched earlier in 2026. The company described the original Haiku-only choice as deliberate: prioritizing speed over analytical depth at launch. The July 23 update acknowledges that session complexity has grown enough to justify offering higher-capability models at the cost of slightly slower turn-taking.

Full-model voice access is not new competitive ground. OpenAI's voice mode already supports the GPT-5.6 line including its reasoning-optimized variants, per TechCrunch's coverage of the Claude update. Anthropic's differentiating bet here is the connector layer: combining capable models with live app actions inside a single session, rather than offering a pure conversation interface. Whether five apps is enough to drive meaningful adoption beyond early adopters remains to be seen.

The move fits a direction visible in Anthropic's recent product work. Claude Cowork, which launched earlier this year, pulled the assistant interface closer to team workflows rather than treating Claude as a standalone chat window users leave to take action elsewhere. Voice mode connectors follow the same logic: reduce the number of app switches required to complete a task.

What to watch next

Connector expansion is the first signal to watch. Gmail, Calendar, Slack, Canva, and Notion cover a sensible core, but the list leaves out project management tools (Asana, Jira), CRM platforms (Salesforce, HubSpot), and most of the category-specific tools enterprise teams depend on. How quickly Anthropic adds to the five-app roster will indicate how much sustained engineering attention the voice-connector surface is receiving versus treatment as a beta feature.

Free-plan limits are the second. The single-connected-tool ceiling means most users who try voice connectors for the first time will hit a wall before they can test the multi-app sequences that make the feature most useful. That friction may keep voice-mode connector adoption concentrated among paid subscribers rather than generating broad discovery through the free tier.

Sources