Several weeks after OpenAI launched its new generation of dialogue models and upgraded the voice function of ChatGPT, Anthropic officially released a major update to the Claude voice mode on July 23. Users can now freely switch between three different model levels: Opus, Sonnet, and Haiku. The voice mode will default to automatically matching the fastest version based on the last model used in text chats.

Upgrading from simple voice to a versatile assistant

image.png

The Claude voice mode was first launched last year, previously driven by the faster but limited Haiku model by default. Although it responded quickly, it was not suitable for handling complex tasks. This upgrade significantly improved performance in long conversations. According to the official statement, the new voice mode can provide targeted feedback on users' communication styles, help users practice customer pitches, and even jointly develop product market research plans. A bigger highlight is the ability to integrate with application ecosystems. The system can directly call third-party applications such as Gmail, Google Calendar, Slack, Canva, and Notion. Users can update meeting schedules, draft emails, or create new documents in Notion through voice commands.

In contrast, although OpenAI has updated the tone and expression style of the voice, it still cannot directly call external tools to collaborate on actual work. In terms of language support, the Claude voice mode now supports ten languages including English, French, German, Japanese, and Korean, and users need to manually switch to the specified language. The new voice mode is now available for Beta testing users on all platforms. Free users can only use the Haiku model and connect to a maximum of one external application. However, Anthropic did not change the underlying voice architecture in this update and did not reveal specific voice technology stacks. There may be no substantial improvements in natural conversation experiences such as interruption handling.