Anthropic has released an update to its voice mode for the Claude chatbot, enhancing its capabilities by incorporating the Sonnet and Opus models for more complex queries. Previously, voice mode operated solely through Haiku, the company’s smallest model, leading to inefficiencies with complicated requests.
The update allows paid users to switch between Haiku, Sonnet, and Opus during conversations. Voice mode will default to the last model used for text chat, ensuring smoother interactions. “Voice mode uses the fastest version of whichever model you’ve selected, so the conversation runs smoothly,” Anthropic stated.
Additionally, the updated voice mode now accesses context from connected applications like Gmail and Slack, contingent on user permission. An Anthropic spokesperson elaborated that Claude’s voice mode operates on a turn-based architecture, meaning it listens, pauses, and then responds, which differs from OpenAI’s fully duplex GPT-Live system that processes speech simultaneously.
This turn-based model may make conversations with Claude feel less natural compared to ChatGPT. Another limitation is that Claude currently cannot automatically detect language changes; users must either explicitly state the switch or manually select the language from the settings menu. However, additional languages, including Indonesian, have been introduced.
“This release is focused on intelligence and tool access,” Anthropic said. “We’re continuing to invest in voice and we’ll have more to share later this year.”
The new voice mode is being rolled out in beta to all users across desktop, mobile apps, and the web client. Users with free accounts will be limited to a single connection, and all prompts will run through Haiku, yet they can still utilize voice mode in all supported languages.
