[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-openai-gpt-live-full-duplex":3,"topics-all":36,"news-related-4d9ec7e0-b5a7-4c7b-b052-e7309c5c225f":55},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"4d9ec7e0-b5a7-4c7b-b052-e7309c5c225f","OpenAI 推出 GPT-Live:全双工语音模型把 ChatGPT 拆成「对话层 + 推理层」","OpenAI 在 7 月 8 日发布 GPT-Live-1 和 GPT-Live-1 mini 两款语音模型,核心变化不是「声音更像人」,而是底层跑通了一套全双工(full-duplex)语音架构:模型边听边说,每秒钟能做多次「该不该继续听\u002F该不该插话\u002F是否要调工具」的决策,把过去 Advanced Voice Mode 上 1.7 秒的串级 ASR-LLM-TTS 延迟压到「边说边回」的连续流。\n\n技术上有两个关键判断。其一是流式对话:旧版基于静音检测判定用户「说完了」,导致任何停顿、背景音都被误识为轮次结束,这也是「对讲机式体验」的根源;GPT-Live 把轮次判定做成连续信号,允许「嗯\u002F好的」这种确认声嵌入主发言,真正像人类会议里的 backchannel。其二是解耦推理:简单问题由 GPT-Live 自身回答,需要查资料或复杂推理时,语音前端异步委托给 GPT-5.5 这类后端大模型,前端继续维持对话流不中断。OpenAI 明确说,未来切换到更新的 frontier 模型时,语音层不用重训——这是一次「语音 UI 与模型智能解耦」的架构押注。\n\n产品层面,GPT-Live-1 已成为 Plus\u002FPro\u002FGo 付费用户的默认语音模型,mini 给免费层;同时新增三层推理档位(Instant\u002FMedium\u002FHigh)和可暂停不打断的能力。对 1.5 亿周活跃语音用户来说,最直观的感受是 ChatGPT 不再「抢话」。\n\n值得讨论的是,这套架构对企业级 voice agent 是真正利好:语音交互的延迟与后端推理能力第一次可以独立优化,客服、导购等场景不必再为「等大模型返回」而把对话节奏打碎。","https:\u002F\u002Fopenai.com\u002Findex\u002Fintroducing-gpt-live\u002F","bd0e0e04-6bcf-4b3e-9a56-62c672308ec9",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"e676a5cf-1f24-472f-a765-86fa21a1bc3c","ai-model",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"baf131c1-687a-49f4-87f6-4dd87c1c692f","gpt",{"id":18,"name":19,"slug":19,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":21,"name":22,"slug":22,"description":13,"color":13},"42e59a88-7795-47dc-a334-ef1e72c24347","openai",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"b94c0bdf-abec-4e27-8f44-a35abac9c853","en","GPT-Live: duplex voice splits ChatGPT into two layers","OpenAI on July 8 released two voice models, GPT-Live-1 and GPT-Live-1 mini. The core change isn't \"voice sounds more human\" but a full-duplex voice architecture running through at the bottom: the model listens and speaks simultaneously, making multiple \"should I keep listening \u002F should I interrupt \u002F should I call a tool\" decisions per second, compressing the previous Advanced Voice Mode's 1.7-second cascaded ASR-LLM-TTS latency into a continuous flow of \"speaking while responding\". There are two key technical choices. First, streaming dialogue: the old version determined the user \"was done speaking\" based on silence detection, causing any pause or background noise to be misrecognized as a turn end — the source of the \"walkie-talkie experience\"; GPT-Live makes turn detection a continuous signal, allowing confirmations like \"mm-hmm\" or \"okay\" to be embedded in the main speech, truly like backchannel in human meetings. Second, decoupled reasoning: simple questions are answered by GPT-Live itself, when research or complex reasoning is needed, the voice frontend asynchronously delegates to a backend model like GPT-5.5, while the frontend continues to maintain conversation flow without interruption. OpenAI clearly states that when switching to a newer frontier model in the future, the voice layer doesn't need to be retrained — this is an architectural bet on \"decoupling voice UI from model intelligence\". On the product side, GPT-Live-1 has become the default voice model for Plus\u002FPro\u002FGo paid users, mini is for the free tier; at the same time three new reasoning tiers (Instant\u002FMedium\u002FHigh) and a pausable-without-interruption capability are added. For 150 million weekly active voice users, the most intuitive feeling is that ChatGPT no longer \"talks over you\". It's worth discussing that this architecture is a real benefit for enterprise voice Agents: voice interaction latency and backend reasoning capability can be optimized independently for the first time, and customer service, sales guidance, and other scenarios no longer need to break the conversation rhythm to \"wait for the big model to return\".","openai-gpt-live-full-duplex","2026-07-09T02:30:00Z","2026-07-09T02:10:33.859741Z","2026-08-19T02:08:40.142862Z",true,"agent",93,[37,46],{"slug":38,"tag_slug":38,"title_zh":39,"title_en":40,"intro_zh":41,"intro_en":42,"id":43,"is_active":33,"created_at":44,"modified_at":45},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":47,"tag_slug":47,"title_zh":48,"title_en":49,"intro_zh":50,"intro_en":51,"id":52,"is_active":33,"created_at":53,"modified_at":54},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":56},[57,62,67,72,77,82],{"id":58,"title":59,"news_slug":60,"published_at":61},"0f466258-290e-4c6b-b471-2169ba6a393f","GPT-Live-1 进 API:全双工语音层 0.05 美元一分钟,推理外包给 GPT-6 Astra","gpt-live-1-api-launch","2026-09-12T17:05:00+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"e12d2e7d-35b7-42d2-b02f-bdcc0a547878","机械臂实测 GPT-6 Astra:19\u002F20 对 8\u002F20 完胜 Fable 5.1,精细插入却全员卡壳","gpt-6-astra-robot-arm-benchmark","2026-09-07T19:13:54+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"390c2437-4e4f-45ec-8270-67c5bfa4fa47","ChatGPT、Claude、Grok、Gemini 罕见同时下线,周四早晨全球 AI 集体失声","chatgpt-claude-grok-gemini-thursday-outage","2026-09-05T06:00:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"b5be4ce8-4a41-461c-9202-148e64fab329","GPT-6 Astra 系统卡:零日自用、对齐升 53%,CoT 可监控性反向下滑","gpt-6-astra-system-card-2026-monitorability","2026-09-04T03:30:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"5bf8fa2d-258e-41d3-bfb6-5c2053433cfd","GPT-6 Astra 正式上线:8 月因安全被暂停的旗舰回来了","gpt-6-astra-launch","2026-09-04T03:12:38+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"9e58d587-3c1b-44c5-ad36-daf23aeb42a2","微软叫停 tokenmaxxing:GitHub Copilot 默认切回 GPT-5.6 Sol,Parikh 设 token 预算","microsoft-token-budget-gpt-5-6-default","2026-09-03T00:30:00+00:00"]