Anthropic this week officially allows paid Claude users to run voice conversations on the Opus and Sonnet models. The upgrade also introduces Claude’s app connectors into voice mode, letting users access Gmail, Google Calendar, Slack, Canva, and Notion through voice. Supported languages expand to 11, including English, French, and more.
Claude Voice Mode Upgrades from Haiku to Opus/Sonnet
According to reports from TechCrunch and 9to5Mac, the core of this upgrade is a tier-level increase in the models that voice mode can run on. Previously, all voice requests were processed through Claude Haiku 4.5 (Anthropic’s smallest and fastest model) to maintain low response times; the new version lets paid users choose Opus or Sonnet voice mode to handle more complex tasks.
Before the announcement was released, TestingCatalog found that the mode selector in the interface existed as a visual control in voice mode for about three weeks, but not until this week did selecting Opus or Sonnet actually change the underlying routing model. Free accounts are still limited to Haiku and a single connected application, but they can still use all supported languages for voice conversations.
App Connectors Enter Voice Mode: Five Tools Support Voice Lookup and Actions
According to reports from TechCrunch and Engadget, Claude voice mode can now access these five connected apps:
Gmail: Write or search emails with voice prompts
Google Calendar: Move or query meetings by voice
Slack: Access Slack message context
Canva: Operate relevant documents
Notion: Create Notion documents with voice prompts
After users grant permissions, voice mode can retrieve context information from the connected apps. Anthropic says this update focuses on “smart analysis and tool access,” with tasks including rehearsing client proposals, getting feedback on communication methods, and conducting product research out loud. TechCrunch notes that, by contrast, the voice mode after OpenAI’s update changed the conversation style but still cannot complete tasks using tools.
Turn-Taking in Claude Voice vs. GPT-Live Listening and Speaking at the Same Time
According to an Anthropic representative’s explanation to TechCrunch, Claude’s voice mode uses a “turn-taking system”: listen first, then think, and then speak; GPT-Live supports listening and speaking at the same time.
TechCrunch also points out that this version update from Anthropic did not change the underlying voice model, and it did not fully explain the technical details of the voice stack. According to TestingCatalog’s testing, Claude still uses text-to-speech technology to generate voice, with voice output coming from ElevenLabs, but the interruption/processing functionality is working normally.
FAQ
Which models are supported after the Claude voice mode upgrade?
According to an Anthropic announcement, paid users can now choose Opus, Sonnet, or Haiku in voice mode, with a mode selector built into the interface; free accounts are still limited to Haiku and a connected application.
Which app connectors does Claude voice mode support?
According to reports from TechCrunch and Engadget, voice mode can now access Gmail, Google Calendar, Slack, Canva, and Notion, and users can use voice prompts to perform actions within these apps.
What are the architectural differences between Claude voice mode and GPT-Live?
According to an Anthropic representative’s explanation, Claude uses a “turn-taking system” (listen first, then speak), while GPT-Live supports listening and speaking at the same time. TechCrunch notes that OpenAI’s updated voice mode changes the conversation style, but it still can’t complete tasks by using tools through voice.
