Claude's voice mode expands to Opus and Sonnet models

When Anthropic launched voice mode last year, it was architected for simplicity: fast responses to quick questions, powered by Claude Haiku, the company's faster but less capable model. But real-world adoption quickly exceeded that design. Users began using voice mode not for casual queries but for deeper work—thinking through business problems, exploring ideas, and requesting actions that only more powerful reasoning could handle. Haiku kept those conversations quick, but not always deep. In response, Anthropic is now rolling out voice mode for Claude Opus and Sonnet, its more advanced models designed for what the company describes as "hard problem-solving." These models can deliver more complex responses and analysis, and crucially, can take action on a user's behalf. A user might dictate a business idea to Opus and ask it to turn the conversation into a one-page pitch, or mention that their train is running late and have the model adjust their calendar appointments accordingly. To make this flexible, Anthropic is allowing users to shift between text and voice mode within a single conversation, and to switch models mid-discussion. A quick question posed to Haiku might spark a deeper idea that a user wants to explore further—they can now seamlessly upgrade to Opus without starting a new thread. The company is also expanding the geographic and linguistic reach of voice mode. Until now, languages beyond English were available only in beta. Voice mode is now live in French, German, Spanish, Hindi, Indonesian, Italian, Japanese, Korean, and Portuguese—removing a significant friction point for non-English speakers who prefer speaking to typing.

Anthropic's original voice mode launch in 2025 centered on speed: quick answers with minimal delay through the lightweight Haiku model. However, the company discovered that real users were applying voice interaction to harder problems—working through business challenges that demanded sustained reasoning and follow-up actions. Haiku, designed to keep conversations quick, was not equipped for this deeper work. By expanding to Opus and Sonnet, Anthropic is responding to demonstrated user behavior rather than anticipated use. The models can now not only generate complex analysis but also take actions: drafting a one-page pitch from a conversation, or reshuffling calendar appointments based on a user's stated need. The mid-conversation model switching feature amplifies this: users can start lightweight and escalate only when necessary, avoiding the cost and latency of running the most powerful model for every query. The expansion of language support—especially the addition of Japanese alongside eight others—signals an intentional move into non-English markets where voice interaction removes typing friction.

No comments yet. Be the first to share your thoughts!