CLAUDE VOICE MODE UPDATE — JULY 24, 2026
● Models: Now Haiku, Sonnet, or Opus — was Haiku only. Defaults to fastest version of your last text model.
● Connectors: Gmail, Slack, Canva, and other connected apps accessible mid-conversation
● Languages: 11 languages — English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Portuguese (Brazilian), Spanish (Latin America/Spain)
● Language switching: Mid-conversation — manually specify the language
● Free tier: Haiku only + one connected app
● Availability: Beta across mobile, desktop, and web
● Architecture: Turn-based (listen → think → respond) — NOT full-duplex
The Core Upgrade — Opus and Sonnet in Voice
Claude's voice mode, which was released last year and powered by the Haiku model, provided quick responses but wasn't well suited for complex work. The company said Thursday that users can choose between Opus, Sonnet, and Haiku models. Voice mode picks the last model people used in the text chat and uses its fastest version by default. This is the most significant capability change to Claude Voice Mode since launch. Haiku was fast but limited — routing a request to talk through an architectural decision or brainstorm a client pitch through a small fast model produced noticeably weaker output. Paid users can switch models mid-conversation through the model picker. Voice mode uses the fastest version of whichever model you've selected, so the conversation runs smoothly, Anthropic notes.
Anthropic said that the new voice mode can help users with longer conversations, including providing feedback on their communication style, talking through a pitch to a client, and brainstorming product market research. These are exactly the tasks that required frontier model intelligence — things Haiku could not handle well in voice mode. A user who paid for Claude Max to access Fable 5 or Sonnet 5 can now actually use those models in voice, rather than getting Haiku-tier responses while paying frontier prices.
Connectors Mid-Conversation — The New Capability
Voice mode can now pull context from connected apps such as Gmail and Slack, as long as you grant Claude permission to do so, and extending its reach into apps like Canva. This is the feature Anthropic teased with the X post: "Voice mode now reaches the tools you've connected mid-conversation." In practice this means: ask Claude to summarise your Gmail inbox while talking to it, have it pull a Slack thread for context during a voice brainstorm, or reference a Canva design mid-conversation without switching to text. The connector access mirrors what Claude already does in text — the same integrations now work in voice.
This is a big difference from OpenAI's voice mode, which updated its conversational style but still isn't able to use different tools to get work done. Tool access in voice is Claude's differentiated capability here — ChatGPT's GPT-Live system can interrupt and respond in real time, but it cannot reach into your Gmail or Slack mid-conversation the way Claude now can.
11 Languages With Mid-Conversation Switching
Earlier this year, Anthropic added multilingual support to Claude's voice mode in beta. Now users can talk in various languages, but they have to manually specify the language. Anthropic supports English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Portuguese (Brazilian), and Spanish (Latin America/Spain). The multilingual expansion that was in beta since June 2026 is now out of beta across all platforms. Language switching mid-conversation works — you ask Claude to switch and it adapts. Manual specification rather than auto-detection is the current limitation: Claude does not automatically detect which language you are speaking and switch accordingly.
The Critical Limitation — Turn-Based, Not Full-Duplex
An Anthropic spokesperson told Engadget voice mode uses a turn-based architecture, so all interactions will see Claude listen to you, pause to think and then respond. It's not fully duplex like OpenAI's new GPT-Live system, which can simultaneously process speech and generate an output. In practice, that should make talking to Claude feel less natural than ChatGPT. This is the honest limitation to state clearly: turn-based means there is a pause between you finishing speaking and Claude starting to respond. Full-duplex (GPT-Live) means the model can start responding while you are still speaking and can be interrupted mid-sentence. For casual conversation, full-duplex feels significantly more natural. For complex task-oriented voice work — the use case Claude is now targeting — the difference matters less.
Claude Voice vs Grok Voice vs ChatGPT Voice — July 2026
| Feature |
Claude Voice |
Grok Voice |
ChatGPT Voice |
| Architecture | Turn-based | Turn-based | Full-duplex (GPT-Live) |
| Model options | Haiku / Sonnet / Opus | Grok 4.5 Aurora | GPT-Live |
| Tool/connector access | Yes — Gmail, Slack, Canva | No | No |
| Live data | No | Live X data | Web search |
| Free voice | Haiku + 1 app | 15 min/day | Limited |
| Languages | 11 | English (primary) | Multiple |
The verdict: Claude wins on tool access (unique — no competitor matches Gmail/Slack in voice) and model quality (Opus/Sonnet now available). ChatGPT wins on conversational naturalness (full-duplex GPT-Live). Grok wins on free voice allocation (15 min/day) and live data access.
Sources: TechCrunch · 9to5Mac · Time News · Engadget · Related: Grok Voice vs ChatGPT Voice → · Claude plan limits →