Anthropic and OpenAI both upgrade voice mode, but in opposite directions
新智元 · wechat · 2026-07-24
Anthropic and OpenAI both rolled out major voice-mode upgrades, but they are taking very different product paths.
Anthropic Claude Voice
- Voice mode now expands from Haiku to the full Opus 4.8 and Sonnet 5 family.
- It can work with Gmail, Calendar, Slack, Google Docs, and Canva through Connectors.
- Users can switch models mid-conversation, and the voice session keeps context with text chats.
- It supports 11 languages, but Chinese is not included.
- Free users get Haiku plus one connector; Opus access requires a paid plan.
OpenAI ChatGPT Voice
- GPT-Live, a new full-duplex voice model, is now on macOS and Windows desktops.
- It uses a two-layer design: a low-latency front end for live conversation and a GPT-5.5 back end for deeper reasoning, web search, and agent tasks.
- OpenAI calls this “delegated reasoning,” letting the assistant keep talking while harder work happens asynchronously.
- Desktop voice can be interrupted, can run multiple agent tasks in parallel, and can use Appshots on macOS to read the active window.
- Plus, Pro, Business, Edu, and Enterprise users can access it now.
The article frames Claude as the more SaaS-connector-driven voice assistant, while ChatGPT aims for a more natural, multitasking desktop copilot that can also control the computer.
More from Models
- Same Echo Maze prompt, three frontier models: all passed visually but shipped the same hidden bug — eyishazyer · 2026-09-11
- Benchmark scores drop from 89% to 19% on new evals — how benchmaxxing breaks leaderboard trust — airesearch12 · 2026-09-11
- ChatGPT tells user their question is too hard and to 'accept dumber answers' — phido3000 · 2026-09-11
- Claude is no longer available for minors as Anthropic rolls out age assurance — Muhammad523 · 2026-09-11
- Developer Building a Unified Leaderboard of All Model Benchmark Scores — airesearch12 · 2026-09-11
- Rumor claims Kimi faked performance by serving Claude; DeepSeek new model surprises in evals — realsohamparekh · 2026-09-11