Anthropic has been offering voice mode for its Claude chatbot since 2025, letting users speak prompts instead of typing them. A July 2026 update made the feature significantly more useful: previously, voice mode only worked through Haiku, Anthropic's least powerful model, to keep latency low. Now it also supports the more capable Sonnet and Opus models, and can pull context from connected apps like Gmail, Google Calendar, and Slack.
Voice mode is available on Claude's mobile apps (Android and iOS), desktop apps, and the web, with Anthropic recommending phone use for the best experience. On mobile, users tap the black waveform icon in the lower right corner (not the adjacent microphone icon), grant microphone permissions if needed, and start talking. Claude responds after each prompt and stays in voice mode until the user taps Stop.
The feature uses a turn-based architecture, meaning Claude listens fully before responding—unlike ChatGPT's newer duplex system, which processes speech and generates output simultaneously. This can cause Claude to mistake a brief pause for the end of a question, so Anthropic suggests asking multi-part questions one at a time.
Users can switch models via the picker at the bottom of the interface: Haiku for fastest responses, Sonnet for everyday tasks, and Opus for complex tasks (Paid accounts only). Voice mode can access third-party apps by mentioning them in the prompt; first-time access requires granting permission.
As of July 2026, voice mode supports 14 languages with some dialects, including English (US, UK, India), Chinese (Simplified), Cantonese, French (Canada), German (Austria), Hindi, Italian, Japanese, Korean, Portuguese (Brazil, Portugal), Russian, Spanish (Latin America), Turkish, and Ukrainian. The chosen language affects voice options—English offers five (Buttery, Airy, Mellow, Glassy, Rounded), while Japanese has only two. Claude cannot detect mid-sentence language switches, so users should state intent or change the language in Settings.
Settings also let users adjust speech speed (Slow, Normal, Fast) and choose between Hands free and Push to talk recording modes, with the former recommended for quiet environments. A paid account is not required for voice mode, but free users are limited to Haiku and Sonnet; Opus requires a subscription. Voice conversations count against usage limits, and free accounts are restricted to a single connection, though all languages are available.
Source: Engadget