Anthropic brings its smartest Claude models to voice mode and plugs them into Gmail and Slack

Anthropic is getting serious about voice, and it’s not just for quick trivia anymore. The company just unlocked its most powerful models—Opus and Sonnet—for voice mode, a significant upgrade from the lightweight Haiku model that previously handled all spoken interactions. This shift acknowledges something Anthropic apparently didn’t see coming: users immediately tried to solve real work problems by talking to Claude, only to hit a wall with Haiku’s limited reasoning depth. As the company put it, those conversations were “quick, but not always deep.”

Sonnet and Opus change that calculus. These are the models Anthropic says are purpose-built for hard problem-solving. The practical upshot is an assistant that can not only discuss a complex issue with you but then act on the conversation—transforming a brainstorming session into a structured one-page pitch or rearranging your calendar when your train is late. You’re also no longer locked into a single mode or model. You can start with a rapid-fire text exchange and switch to a spoken deep dive with Opus without breaking your flow, which feels like the kind of flexibility that actually fits how people work.

Anthropic is also pushing Claude’s voice into the places where work happens by integrating with Gmail, Slack, and Canva. That’s a direct shot at making the assistant contextual rather than a standalone app you have to consciously visit. The rollout also expands language support beyond English and beta tests to nine new languages, including French, German, Spanish, Japanese, and Portuguese. It’s a move that signals Anthropic isn’t just building a research toy—it’s building a tool that needs to function in multilingual, multi-app work environments right now.