STORY RECORD
Anthropic arms Claude voice with reasoning and tools
Anthropic upgraded Claude voice mode on July 23 from the lightweight Haiku-only system to support Opus, Sonnet, and Haiku models, while also giving voice conversations real-time access to connected tools like Gmail, Slack, Calendar, Canva, and Notion.
The update rolls out as a public beta on mobile, desktop, and web with 10 languages. Free accounts stay locked to Haiku with one connected app. The underlying TTS stack is unchanged — no native audio model or interruption improvements.
The timing matters: OpenAI shipped GPT-Live this week with better conversation flow but zero external tool access. Anthropic is betting that being able to say 'reschedule my meeting and send the notes to Nick in Slack' is more valuable than a smoother back-and-forth.
Robert Bye, Anthropic PM, confirmed the update. The Verge reports user behavior drove the shift — people were already using voice for 'real business problems' Haiku couldn't handle.
TestingCatalog found a hidden model selector in the UI had been defaulting to Haiku for three weeks before the switch.
Why It Matters
This marks a strategic fork in the voice AI race. OpenAI just shipped GPT-Live with impressive conversational polish — better interruption handling, more natural flow — but it still cannot touch external tools.
Anthropic's bet is that being able to act on what you say matters more than how smoothly you say it.
If a user can dictate 'move my Tuesday meeting, draft an email about the change, and summarize what I missed in Slack' and have Claude execute all three, the voice assistant transforms from a chat interface into a hands-free productivity agent.
The update also extends Anthropic's aggressive July feature cadence — multiplayer Claude Code, in-app browser, managed agents, Fable 5 — into the voice modality. For power users and builders, this could redefine what a voice assistant is supposed to do.
The open question: will users accept a less polished voice experience in exchange for real-world action capability?
The Facts
10Claude voice mode previously only ran on the Haiku model; the July 23 update adds support for Opus and Sonnet, Anthropic's more capable reasoning models.
strong · confidence 0.95
Voice mode can now access connected tools mid-conversation, including Gmail, Google Calendar, Slack, Canva, and Notion, enabling actions like rescheduling meetings, drafting emails, and summarizing Slack messages.
strong · confidence 0.9
Users can switch between text and voice mode mid-conversation and change models on the fly without losing context.
moderate · confidence 0.8
Voice mode now supports 10 languages including English, French, German, Spanish, Hindi, Indonesian, Italian, Japanese, Korean, and Portuguese (Brazilian).
strong · confidence 0.85
The update rolled out July 23, 2026 as a public beta across mobile, desktop, and web, available on all plan tiers.
strong · confidence 0.9
Free accounts are restricted to the Haiku model with only one connected app.
moderate · confidence 0.7
Anthropic did not change the underlying TTS voice model or add native audio model capabilities; the update is built on the same text-to-speech stack without the interruption-handling or conversational-flow improvements OpenAI shipped with GPT-Live.
moderate · confidence 0.8
Robert Bye, Anthropic Product Manager, confirmed the update with example use cases: 'Summarise what I missed in Slack' and 'Share a tl;dr of my new project PRD Google Doc, and send it to Nick in Slack.'
strong · confidence 0.9
TestingCatalog independently detected a hidden model selector in the Claude UI for roughly three weeks that defaulted to Haiku before the July 23 switch enabled Opus and Sonnet responses in voice mode.
moderate · confidence 0.8
The Verge independently confirmed the upgrade, reporting that Anthropic's blog post said users were using voice mode for 'real business problems' that Haiku was not designed for.
strong · confidence 0.9
Still Open
1- ContradictionContradiction note is limited to snapshotted evidence.contradicted