Updated Claude Voice Mode (9 min read)

Anthropic has released an update that significantly improves the performance of Claude voice mode. Previously, the lightweight Haiku model was mainly used to reduce latency, but now users can select Sonnet or Opus models to handle more complex requests.

The key updates are as follows:

• Model selection freedom: You can switch between Haiku, Sonnet, and Opus mid-conversation via the model picker.
• App integration features: With permission granted, you can pull in context from external apps such as Gmail and Slack.
• Expanded language support: Additional language support has been expanded, including Indonesian.

However, unlike OpenAI's GPT-Live, Claude's voice mode uses a turn-based architecture that listens to the user, thinks, and then responds. Because of this, it may feel less natural than ChatGPT in terms of real-time two-way conversation, and it has the constraint of not being able to automatically detect language changes mid-conversation, requiring users to set it directly or state it verbally.

This update is being rolled out as a beta version via desktop, mobile apps, and web clients. Free account users are limited to the Haiku model, but can still have voice conversations in all supported languages.