# Claude Voice Mode Gets Opus, Sonnet, and App Integrations

Anthropic upgraded Claude Voice to Opus/Sonnet models and added Gmail, Slack, and Notion integrations, turning the quick-answer tool into a voice assistant that does real work across your apps.

By Neomanex. Updated 2026-10-07.

**Verdict: This isn't a new voice model — it's a capability upgrade that turns Claude Voice from a quick-response tool into a work-doing assistant. The app integrations are the real differentiator: OpenAI's GPT-Live still can't touch your calendar.**

Anthropic announced the update on July 23, 2026 (Anthropic, 2026), upgrading Claude's voice mode — which launched last year powered exclusively by the Haiku model — to support the full Opus and Sonnet model families. Voice mode now defaults to the last model used in text chat and runs its fastest variant by default. A model picker lets users swap between Haiku, Sonnet, and Opus mid-conversation (Anthropic, 2026).

## What happened

Until this update, [Claude Opus 4.8](/models/claude-4-opus) and [Claude Sonnet 5](/models/claude-sonnet-5) were text-only. Every voice session previously routed through Haiku regardless of which model appeared in the picker. TestingCatalog reported that the model picker was present but non-functional for roughly three weeks before this release, with all sessions still going to Haiku 4.5 (via Cryptopolitan, 2026).

Three things changed on July 23 (Anthropic, 2026):

1. **Model routing is now real.** Voice sessions use whichever model the user last selected in text chat, running the fastest variant of that model family to keep latency manageable (Anthropic, 2026).
2. **App connectors entered voice mode.** Claude Voice can now reach into Gmail, Google Calendar, Slack, and Canva (Anthropic, 2026), plus Notion (Notion is included per TechCrunch and The Verge but was not listed in Anthropic's own blog post) — a spoken request can reschedule a meeting, draft and send an email, or create a document without touching a keyboard.
3. **10-language support** is now official: English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Portuguese (Brazilian), and Spanish (Latin America/Spain). Users must manually specify the language — there is no automatic language detection yet (Anthropic, 2026).

The tiering is straightforward (Anthropic, 2026):

| Plan | Models | Connected apps |
|------|--------|---------------|
| Free | Haiku only | 1 app |
| Pro / Max / Team / Enterprise | Opus, Sonnet, Haiku | Multiple apps |

Claude Fable 5 — Anthropic's most capable widely available model — is not included in voice mode (Unite.AI, 2026).

## Why it matters

The app integration gap is the story. OpenAI launched GPT-Live on July 8 with a new full-duplex voice architecture — the AI can listen and speak simultaneously, creating a more natural conversational feel. But GPT-Live cannot use external tools. It talks beautifully, but it can't move your calendar appointment.

Anthropic took the opposite bet: keep the existing turn-based voice architecture (Claude listens, thinks, then responds) and layer stronger reasoning models plus tool access on top. The voice stack is turn-based, not full-duplex (Anthropic, 2026). The TTS output is reportedly powered by ElevenLabs (Cryptopolitan, 2026), though Anthropic has not publicly confirmed its voice stack provider.

| Capability | Claude Voice (July 2026) | GPT-Live (July 2026) |
|------------|--------------------------|----------------------|
| Voice architecture | Turn-based (TTS) | Full-duplex (speech-native) |
| Supported models | Opus, Sonnet, Haiku | GPT-Live models |
| App integrations | Gmail, Calendar, Slack, Canva, Notion | None |
| Languages | 10 (manual selection) | Multilingual |
| Free tier | Haiku, 1 app | Basic access |

The practical tradeoff: Claude Voice feels less natural conversationally but can do more actual work. For enterprise users who have already deployed Claude across their organizations — Deloitte with 470,000 employees and the Alberta government spanning 27 ministries (figures per Cryptopolitan, 2026) — the ability to query emails, calendars, and Slack messages by voice makes the assistant materially more useful inside existing workflows.

## What changes for you

**If you're a paid Claude subscriber**, voice mode now inherits the intelligence of your chosen model (Anthropic, 2026). Use Sonnet 5 for agentic coding discussions or Opus 4.8 for complex analytical reasoning — by voice. Connect your Gmail and Slack accounts in settings, then ask Claude to summarize your inbox, reschedule conflicts, or draft replies.

**If you're on the free tier**, you get Haiku voice with one connected app (Anthropic, 2026). It's still a meaningful upgrade from text-only interaction, but the real power — multi-model voice with full app connectivity — requires a paid plan.

**If you're comparing to ChatGPT**, the choice now splits cleanly: GPT-Live for conversations that feel natural, Claude Voice for conversations that get work done. Neither does both yet.

**Expect voice quality to stay the same.** Anthropic didn't change the underlying voice model with this release. Interruption handling, turn-taking, and naturalness are unchanged from before. The upgrade is about what the model behind the voice can understand and do — not how smoothly it talks.

## FAQ

**Is this a new voice model?** No. The voice stack is unchanged — same turn-based architecture (Anthropic, 2026), same TTS provider. The upgrade routes your spoken requests through Opus or Sonnet instead of Haiku, giving the AI more reasoning power behind the same voice output.

**Why can't Claude Voice do full-duplex like GPT-Live?** Anthropic chose turn-based architecture (listen → think → respond) rather than simultaneous listening and speaking (Anthropic, 2026). This prioritizes response quality over conversational fluidity. The tradeoff is deliberate — stronger reasoning over smoother chatting.

**Can I use Claude Fable 5 in voice mode?** No. Only Opus, Sonnet, and Haiku are available. Anthropic hasn't said when or if Fable 5 will come to voice.

Canonical: https://neomanex.com/news/claude-voice-mode-opus-sonnet-app-integrations