{"slug": "anthropic-upgrades-claude-voice-mode-to-run-opus-and-sonnet-while-automating-and", "title": "Anthropic upgrades Claude voice mode to run Opus and Sonnet while automating Gmail, Calendar, and Slack", "summary": "Anthropic on July 23 upgraded Claude's voice mode to support Sonnet and Opus models and integrated it with Gmail, Google Calendar, Google Docs, Slack, Canva, and Notion, allowing users to perform tasks like checking calendars, reading emails, drafting Slack messages, and starting documents via voice. The update fixes the previous limitation of routing voice queries through the smaller Haiku model, but the real significance lies in the connected-app layer that enables spoken requests to interact with workplace software, raising governance questions about saved transcripts of voice conversations.", "body_md": "*Anthropic's July 23 update gives Claude voice mode stronger models and connected apps, but the real test is whether companies are ready for workers to talk an AI into acting on their inboxes and calendars.*\n\nUntil this week, Claude's voice mode was stuck on Haiku. Fast, yes. Built for deeper work, not really. Anthropic changed that on July 23 by adding Sonnet and Opus to voice conversations and tying those conversations more tightly to connected apps including Gmail, Google Calendar, Google Docs, Slack, Canva, and Notion. You can now ask Claude out loud to check your calendar, read relevant email context, draft a Slack message, or start a document. That is the useful part. The voice is no longer just an input method.\n\nThe model change is worth pausing on because it fixes Claude voice mode's most obvious weakness. As Engadget reported, the feature previously routed voice queries through Haiku, Anthropic's smallest model, partly to keep latency down. Now paid users can use Sonnet or Opus in voice mode, and the tool can default to the model they last used in text chat. The Verge also reported that users can switch among models mid-conversation. If you're paying for Claude because you want Opus-level reasoning, it never made much sense that speaking to Claude forced you back to Haiku.\n\nOpenAI still has the more natural voice product. TechCrunch's early review of Advanced Voice Mode described it as more like talking with a phone than using one, and OpenAI's own help pages now describe GPT-Live-1 as able to listen and speak at the same time. Claude is taking a different route. It is less about perfect conversational flow and more about letting a spoken request touch the software where your work already lives.\n\n## The useful part is the app layer\n\nThe connected-app layer is where this update becomes more than a model picker. Claude can pull context from tools you have granted it permission to use, so a spoken request about this afternoon's meetings can draw from both Google Calendar and Gmail rather than forcing you to open each app yourself. Free users remain limited to Haiku and one connected app, while paid plans get access to more capable models and multiple connected apps. That split is blunt, but it gives Anthropic a clear upgrade path. Pay, and voice becomes a work surface rather than a dictation feature.\n\nThere are limits. The rollout supports English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Portuguese, and Spanish, with Portuguese and Spanish split into regional variants in some coverage. Claude's own help center says voice mode supports hands-free conversations and push-to-talk, and it also confirms that textual transcripts of voice conversations are saved to chat history just like text chats. Keep that detail in your head. It matters more than the language list.\n\nA spoken request feels casual. A saved transcript is not casual. If an employee asks Claude to read an email aloud, summarize a calendar conflict, and draft a reply in Slack, the conversation can leave a written trail in the same place a typed Claude chat would. That is not automatically a problem, but it is a governance question most companies haven't answered. Voice lowers the friction for sharing context, and lower friction changes behavior. People will say things they wouldn't have taken the time to type.\n\nClaude voice mode is also not available everywhere Anthropic works. The update does not extend into Claude Cowork or Claude Code, according to coverage of the rollout, which keeps voice-triggered connected-app work separate from Anthropic's developer-facing products for now. That boundary is sensible. If you're letting an AI act across Gmail and Slack, you want permissioning and audit trails settled before the same pattern moves deeper into code and production workflows.\n\n## The race is about work, not sound\n\nAnthropic is not winning on vocal polish. ChatGPT has the stronger claim there, and OpenAI has moved quickly with GPT-Live-1, background conversations, and voice settings for intelligence levels. So don't frame this as a simple voice assistant race. The better comparison is workplace automation, the kind that usually arrives through admin panels, procurement reviews, and long internal arguments about who is allowed to automate what.\n\nFrankly, many companies are not ready for the everyday version of this. An employee typing a bad instruction may leave a draft sitting in a compose window. An employee speaking a bad instruction to an AI with app access may move faster than the review process around it. That is exactly why this update matters. It makes capable models easier to use in the places where workers already make quick decisions.\n\nThe model-tier move is still smart. Giving paid users Sonnet and Opus through voice creates a real reason to upgrade beyond abstract promises about better reasoning. That matters. The open question is whether Claude's workflow reach can feel reliable when several app actions are chained in a single spoken session. If it can, Anthropic has made voice less of a demo and more of a daily work interface.\n\n**Also read:** [Midjourney buys Co-Star to seed a consumer app empire it never needed to build before](https://startupfortune.com/midjourney-buys-co-star-to-seed-a-consumer-app-empire-it-never-needed-to-build-before/) • [CATL's energy storage business grew 88% in six months as AI data centers become its next major market](https://startupfortune.com/catls-energy-storage-business-grew-88-in-six-months-as-ai-data-centers-become-its-next-major-market/) • [Congress introduces an AI kill switch bill after an OpenAI model hacked its way out of a sandbox](https://startupfortune.com/congress-introduces-an-ai-kill-switch-bill-after-an-openai-model-hacked-its-way-out-of-a-sandbox/)", "url": "https://wpnews.pro/news/anthropic-upgrades-claude-voice-mode-to-run-opus-and-sonnet-while-automating-and", "canonical_source": "https://startupfortune.com/anthropic-upgrades-claude-voice-mode-to-run-opus-and-sonnet-while-automating-gmail-calendar-and-slack/", "published_at": "2026-07-24 15:12:51+00:00", "updated_at": "2026-07-24 15:24:13.523223+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-products", "ai-tools", "large-language-models"], "entities": ["Anthropic", "Claude", "Sonnet", "Opus", "Haiku", "Gmail", "Google Calendar", "Slack"], "alternates": {"html": "https://wpnews.pro/news/anthropic-upgrades-claude-voice-mode-to-run-opus-and-sonnet-while-automating-and", "markdown": "https://wpnews.pro/news/anthropic-upgrades-claude-voice-mode-to-run-opus-and-sonnet-while-automating-and.md", "text": "https://wpnews.pro/news/anthropic-upgrades-claude-voice-mode-to-run-opus-and-sonnet-while-automating-and.txt", "jsonld": "https://wpnews.pro/news/anthropic-upgrades-claude-voice-mode-to-run-opus-and-sonnet-while-automating-and.jsonld"}}