Claude's Voice Mode Gets a Major Upgrade: You Can Now Choose Your Model Instead of Being Stuck with Haiku

When "talking to AI" gets real work done, not just casual chat

On July 23, 2026, Anthropic announced a significant update to Claude's Voice Mode. The main change is that users can now choose their own model among Opus, Sonnet and Haiku, where the voice system previously ran on Haiku alone.

This is not the first launch of Voice Mode (the feature has existed since 2025). It is an overhaul that makes it usable for more serious work, especially complex tasks.

The old problem: fast but shallow

Until now, Claude's Voice Mode ran on Haiku, the smallest model. The advantage is fast responses, but once the work gets complex, such as analysis, in-depth advice or long conversations, it starts to run out of depth.

This update addresses that problem directly.

Choose the brain that fits the task

You can now choose which model your Claude voice runs on:

Haiku — fast, suited to short questions that need quick answers
Sonnet — a balance of speed and capability, suited to general tasks
Opus — the most capable, suited to work that needs deep thinking and long conversations

Conveniently, the system automatically picks up the latest model you used in text chat, so you do not have to set it up again each time.

Where it beats the competition: speak a command and it does the work

The highlight for working professionals is that Claude's Voice Mode connects to real work apps, including Gmail, Google Calendar, Slack, Canva and Notion.

This means you can give one spoken instruction and have Claude:

Reschedule a meeting in Google Calendar
Draft an email in Gmail
Create a document in Notion

This is a key difference from ChatGPT's Voice Mode, which has a smoother conversational style but still cannot use tools to carry out work for you.

What you can use it for in real working life

Anthropic gives examples of uses that suit long conversations, such as:

Rehearsing a client pitch — talk with Claude as if you were practicing a real presentation
Getting feedback on how you communicate — ask it to analyze your speaking style
Brainstorming market research ideas — keep developing ideas while commuting or driving

What to know before you try it

The feature is still in beta, rolling out gradually to everyone on every platform (mobile, desktop and web)
It supports several languages (English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Portuguese and Spanish), but does not yet support Thai, and you have to select the language yourself

Summary

This update moves Claude's Voice Mode from "an assistant that answers questions by voice" to "an assistant that does the work when you speak a command." Being able to choose the model, together with connections to work apps, makes it a better fit for professionals and business owners who want to use AI hands-free.

Although Thai is not supported yet, it is a clear signal that AI is moving toward fully voice-driven work, and worth keeping an eye on.

Source: Anthropic