Claude Voice Mode Now Supports Opus and Sonnet, Connected Apps, and More Languages in Beta

Anthropic rolled out a major update to Claude’s voice mode on July 23, 2026²³. Voice mode previously ran only on Haiku, the company’s lightweight model tuned for fast responses — but with this update it can use the more capable Opus and Sonnet models as well³. The update also brings connected-app support (Gmail, Slack, and more) into voice conversations, plus beta support for additional languages including Japanese¹.

Voice assistants have long carried an implicit constraint: quick answers, shallow thinking. With the same intelligence as text chat now available by voice, working through genuinely complex questions while commuting or between tasks becomes a realistic workflow.

The Same Models as Text Chat

According to Anthropic’s help center, voice mode starts with the model you last used in text chat (Sonnet, Opus, or Haiku) and automatically switches to the latest generation of that model¹. You can also switch models mid-conversation using the model selector¹ — casual questions on fast Haiku, harder problems on Opus, all within a single session.

Note that Claude Fable (Fable 5), Anthropic’s top-tier model, is not currently available in voice mode¹. And according to TechCrunch, free-plan users are restricted to Haiku in voice mode².

Operating Connected Apps by Voice

The second pillar of the update is app integration. Voice mode can now use the tools you’ve connected to Claude. The official help page lists Gmail, Google Calendar, Google Docs, and Slack as examples¹, and reporting also mentions Canva². Requests like “catch me up on my email,” “check my calendar before the next meeting,” or “summarize this thread” can now be handled entirely by voice, without looking at a screen¹.

This moves voice mode a step beyond question-answering toward an AI agent-style assistant that calls external tools to do work on your behalf. The number of connectable tools varies by plan: one on the Free plan, more on paid plans¹.

Japanese and Other Languages in Beta, on All Plans and Platforms

Language support has expanded too. Support for languages other than English is in beta¹; 9to5Mac lists Japanese, French, German, Hindi, Indonesian, Italian, Korean, Brazilian Portuguese, and Spanish (Latin America and Spain)³. You can set a preferred language in settings or simply ask Claude to switch languages mid-conversation¹.

Voice mode is available on Claude Mobile (iOS and Android), Claude Desktop, and the web, across all plans — Free, Pro, Max, Team, and Enterprise¹. Anthropic says it is built to work best from your phone¹. Dictation — converting speech to text for typed prompts — remains a separate feature, with voice mode positioned as a full two-way conversation¹.

The Backstory: The Race to Upgrade Voice Assistants

Voice has been a busy front this summer. OpenAI shipped GPT-Live, a full-duplex voice model that listens while it speaks, overhauling ChatGPT Voice in July. Apple opened its conversational new Siri to public beta testers in iOS 27. Where OpenAI focused on making the voice model itself more natural and responsive, Anthropic’s update swaps in smarter models behind the voice and connects it to work tools — TechCrunch notes the release includes no changes to the underlying voice model².

Hands-Free Work Gets Practical

For readers who use Claude at work, the implications are straightforward. Voice mode used to be a research-and-chat feature; with capable models and app connections, a single conversation can now cover triaging your inbox on the commute and reviewing your schedule and relevant threads right before a meeting. The more hands-busy your day, the bigger the benefit.

A few caveats. Non-English support, including Japanese, is still in beta, so recognition and speech quality have room to improve. Free-plan users are limited to Haiku (per reporting²) and one connected tool, which effectively makes advanced voice work a paid-plan feature. A practical first step: connect the one tool you use most, try it by voice on your current plan, and see whether it fits your workflow.