[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-grok-46-matches-top-models-spacex-worker-data-zh":3,"article-related-grok-46-matches-top-models-spacex-worker-data-zh":32,"series-industry-4ae3e0e2-d73f-4b6b-b336-bbb4678f3d90":77},{"id":4,"slug":5,"title":6,"content":7,"summary":8,"source":9,"source_url":10,"author":11,"image_url":12,"cover_image":12,"category":13,"language":14,"translated_content":11,"related_article_id":15,"keywords":16,"key_takeaways":24,"views":29,"created_at":30,"published_at":31,"topic_cluster_id":11},"4ae3e0e2-d73f-4b6b-b336-bbb4678f3d90","grok-46-matches-top-models-spacex-worker-data-zh","Grok 4.6 追上頂尖模型，但資料治理更關鍵","\u003Cp>\u003Ca href=\"\u002Fnews\u002Fgrok-46-frontier-intelligence-cost-efficiency-zh\">Grok\u003C\u002Fa> 4.6 到底值不值得看，SpaceX 又想拿員工哪些資料來訓練？\u003C\u002Fp>\u003Cp data-speakable=\"summary\">Grok 4.6 的測試表現已追平頂尖模型，SpaceX 則計畫用員工\u003Ca href=\"\u002Fnews\u002Fchatgpt-mac-computer-history-memory-zh\">工作\u003C\u002Fa>資料訓練未來版本。\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>項目\u003C\u002Fth>\u003Cth>AI Index\u003C\u002Fth>\u003Cth>Input price\u003C\u002Fth>\u003Cth>Output price\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>Grok 4.6\u003C\u002Ftd>\u003Ctd>61\u003C\u002Ftd>\u003Ctd>$2 \u002F 1M tokens\u003C\u002Ftd>\u003Ctd>$6 \u002F 1M tokens\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>\u003Ca href=\"https:\u002F\u002Fopenai.com\u002F\">OpenAI\u003C\u002Fa> GPT-5.6 Sol\u003C\u002Ftd>\u003Ctd>61\u003C\u002Ftd>\u003Ctd>$5 \u002F 1M tokens\u003C\u002Ftd>\u003Ctd>$30 \u002F 1M tokens\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>\u003Ca href=\"https:\u002F\u002Fwww.anthropic.com\u002F\">Anthropic\u003C\u002Fa> Claude Fable 5\u003C\u002Ftd>\u003Ctd>62\u003C\u002Ftd>\u003Ctd>$5 \u002F 1M tokens\u003C\u002Ftd>\u003Ctd>n\u002Fa\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>Claude Opus 5\u003C\u002Ftd>\u003Ctd>n\u002Fa\u003C\u002Ftd>\u003Ctd>$5 \u002F 1M tokens\u003C\u002Ftd>\u003Ctd>$25 \u002F 1M tokens\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>1. Grok 4.6 的升級路線\u003C\u002Fh2>\u003Cp>\u003Ca href=\"https:\u002F\u002Fx.ai\u002F\">Grok\u003C\u002Fa> 4.6 的重點，不是把底座做得更大，而是把訓練做得更細。公司表示，它沿用 Grok 4.5 的 1.5 兆參數 V9 基礎，再靠更長的補充訓練、強化監督式微調與更多強化學習拉高表現。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786885369221-z6iq.png\" alt=\"Grok 4.6 追上頂尖模型，但資料治理更關鍵\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>這代表性能提升不一定要重做預訓練。對實務團隊來說，這種路線通常更快，也更容易把模型往特定任務推進，例如程式優化、網頁開發與 CAD 工作。\u003C\u002Fp>\u003Cul>\u003Cli>底座：1.5T 參數 V9\u003C\u002Fli>\u003Cli>訓練方式：SFT、RL、trajectory filtering\u003C\u002Fli>\u003Cli>發布時間：2026\u002F08\u002F12\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>2. 61 分代表什麼\u003C\u002Fh2>\u003Cp>在 Artificial Analysis 的測試裡，Grok 4.6 拿到 Intelligence Index 61，明顯高於 Grok 4.5 的 56，也大幅超過 Grok 4.3 的 38。這讓它和 GPT-5.6 Sol 打平，只比 \u003Ca href=\"\u002Ftag\u002Fclaude\">Claude\u003C\u002Fa> Fable 5 低 1 分。\u003C\u002Fp>\u003Cp>更值得注意的是\u003Ca href=\"\u002Fnews\u002Flong-horizon-agents-need-harnesses-first-zh\">代理\u003C\u002Fa>型任務表現。它在 GDPval-AA v2 拿到 Elo 1753，落在 Claude Opus 5 之後，但和 Claude Fable 5、Qwen3.8 Max 同一梯隊。若你在意模型排名，這已經不是「能用」而是「可選」等級。\u003C\u002Fp>\u003Cul>\u003Cli>AI Index：61\u003C\u002Fli>\u003Cli>Grok 4.5：56\u003C\u002Fli>\u003Cli>Grok 4.3：38\u003C\u002Fli>\u003Cli>GDPval-AA v2：Elo 1753\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>3. 價格與效率的實際差距\u003C\u002Fh2>\u003Cp>Grok 4.6 的吸引力不只在分數。Artificial Analysis 指出，它完成知識工作任務平均約 53 回合、0.5 億 tokens 輸入；Claude Opus 5 則大約要 103 回合、2.0 億 tokens。回合更少，通常意味著編排成本更低。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786885369045-62cy.png\" alt=\"Grok 4.6 追上頂尖模型，但資料治理更關鍵\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>價格也更有感。Grok 4.6 維持每百萬 input tokens $2、output tokens $6，明顯低於 \u003Ca href=\"\u002Ftag\u002Fanthropic\">Anthropic\u003C\u002Fa> 與 \u003Ca href=\"\u002Ftag\u002Fopenai\">OpenAI\u003C\u002Fa> 的同類標價。若你的產品流量大、API 成本敏感，這一項很關鍵。\u003C\u002Fp>\u003Ccode>Grok 4.6：$2 in \u002F $6 out per 1M tokens\u003Cbr>Claude Opus 5：$5 in \u002F $25 out\u003Cbr>GPT-5.6 Sol：$5 in \u002F $30 out\u003C\u002Fcode>\u003Ch2>4. SpaceX 想拿什麼資料訓練\u003C\u002Fh2>\u003Cp>\u003Ca href=\"https:\u002F\u002Fwww.spacex.com\u002F\">SpaceX\u003C\u002Fa> 的說法比模型本身更敏感。Elon Musk 表示，公司想把「SpaceX 的全部資訊」用來訓練未來 Grok 版本，包含員工想法與產出，目標是把工程師的判斷能力傳給模型。\u003C\u002Fp>\u003Cp>但目前沒有公開資料範圍、蒐集方式，也沒有說明員工能否拒絕。對\u003Ca href=\"\u002Ftag\u002F企業-ai\">企業 AI\u003C\u002Fa> 來說，最有價值的往往不是文件本身，而是人怎麼做決策、怎麼修正錯誤、怎麼完成多步驟工作。\u003C\u002Fp>\u003Cul>\u003Cli>員工規模：約 14,000 至 15,000 人\u003C\u002Fli>\u003Cli>未公布 opt-out 機制\u003C\u002Fli>\u003Cli>未公開資料來源與保護措施\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>5. Meta 的前例已經說明風險\u003C\u002Fh2>\u003Cp>這類計畫不是沒有前車之鑑。\u003Ca href=\"\u002Ftag\u002Fmeta\">Meta\u003C\u002Fa> 的 Model Capability Initiative 曾嘗試記錄員工電腦活動作為 AI 訓練資料，後來引發反彈，還出現資料事件，讓私人對話、績效資料與逐字稿被跨部門看見。\u003C\u002Fp>\u003Cp>那個案例提醒一件事：如果要用工作行為訓練模型，範圍、同意與存取控制都要先定清楚。否則資料一旦進系統，風險就不再只是理論。\u003C\u002Fp>\u003Cul>\u003Cli>啟動時間：2026 年 4 月\u003C\u002Fli>\u003Cli>事件等級：SEV 2\u003C\u002Fli>\u003Cli>狀態：截至 2026\u002F08\u002F12 仍暫停\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>哪種適合你\u003C\u002Fh2>\u003Cp>如果你最在意 API 成本與效率，Grok 4.6 是這份名單裡最值得盯的選項。若你看重極限排名，Claude Fable 5 和 Claude Opus 5 仍在前段班。若你關心的是公司如何使用員工資料，那麼 SpaceX 的資料政策比模型分數更需要先看懂。\u003C\u002Fp>\u003Cp>簡單說，Grok 4.6 是一個「便宜且夠強」的新選擇，但 SpaceX 的員工資料計畫，會決定它在企業內部能走多遠。\u003C\u002Fp>","4 個重點看懂 Grok 4.6 的 61 分表現、$2\u002F$6 定價優勢，以及 SpaceX 想拿員工工作資料訓練模型的風險。","www.techtimes.com","https:\u002F\u002Fwww.techtimes.com\u002Farticles\u002F324156\u002F20260812\u002Fgrok-46-arrives-spacex-claims-all-employee-work-ai-training-material.htm",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786885369221-z6iq.png","industry","zh","60de465c-c7f2-4aac-9df0-49338ed86bcc",[17,18,19,20,21,22,23],"Grok 4.6","SpaceX","worker data","AI benchmark","token pricing","Anthropic","OpenAI",[25,26,27,28],"Grok 4.6 以 61 分追平 GPT-5.6 Sol，價格卻更低。","它的提升主要來自後訓練與強化學習，不是更大的底座。","SpaceX 想用員工工作資料訓練模型，但資料範圍與同意機制未公開。","若重視成本與效率可看 Grok 4.6，若重視治理風險則要先看資料政策。",1,"2026-08-16T13:02:21.292285+00:00","2026-08-16T13:02:21.271+00:00",{"tags":33,"relatedLang":36,"relatedPosts":40},[34],{"name":20,"slug":35},"ai-benchmark",{"id":15,"slug":37,"title":38,"language":39},"grok-46-matches-top-models-spacex-worker-data-en","Grok 4.6 matches top models as SpaceX eyes worker data","en",[41,47,53,59,65,71],{"id":42,"slug":43,"title":44,"cover_image":45,"image_url":45,"created_at":46,"category":13},"0edc135b-6fab-470c-bbde-82070ed58a1e","grok-4-6-cheaper-frontier-ai-builders-zh","Grok 4.6 讓前沿 AI 更便宜了","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786887172207-w42s.png","2026-08-16T13:32:22.361586+00:00",{"id":48,"slug":49,"title":50,"cover_image":51,"image_url":51,"created_at":52,"category":13},"a29e54e2-c849-4fe7-97a7-fc17512a4db1","wall-street-real-assets-blockchains-not-just-pilots-zh","華爾街應把真實資產上鏈，而不是只做試點","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786860162975-k8vf.png","2026-08-16T06:02:18.232783+00:00",{"id":54,"slug":55,"title":56,"cover_image":57,"image_url":57,"created_at":58,"category":13},"f5cf0685-7490-43ec-be14-71be76a0270b","workbuddy-jineng-mcp-zhuanjia-qubie-zh","WorkBuddy 里技能、MCP、专家怎么分工","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786824176035-4j5r.png","2026-08-15T20:02:29.690908+00:00",{"id":60,"slug":61,"title":62,"cover_image":63,"image_url":63,"created_at":64,"category":13},"54ada8df-3058-4dd1-94ef-e7e78f611e22","glm-5-3-coding-gains-post-training-zh","GLM-5.3 編碼提升，重點在後訓練","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786773771566-dghc.png","2026-08-15T06:02:18.29616+00:00",{"id":66,"slug":67,"title":68,"cover_image":69,"image_url":69,"created_at":70,"category":13},"26c83d04-4b98-4e82-95f0-1e94e19d2c92","anthropic-data-center-push-big-capital-backing-zh","Anthropic 50 億美元數據中心背後的 5 個資本動作","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786752165620-j8sc.png","2026-08-15T00:02:21.073927+00:00",{"id":72,"slug":73,"title":74,"cover_image":75,"image_url":75,"created_at":76,"category":13},"057e9280-7f45-499f-85c7-bd678f90c812","dailyarxiv-arxiv-keyword-paper-feed-zh","DailyArxiv 把 arXiv 關鍵字變成每日論文流","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786716168076-7w4o.png","2026-08-14T14:02:22.664653+00:00",[78,83,88,93,98,103,108,113,118,123],{"id":79,"slug":80,"title":81,"created_at":82},"ee073da7-28b3-4752-a319-5a501459fb87","ai-in-2026-what-actually-matters-now-zh","2026 AI 真正重要的事","2026-03-26T07:09:12.008134+00:00",{"id":84,"slug":85,"title":86,"created_at":87},"83bd1795-8548-44c9-9a7e-de50a0923f71","trump-ai-framework-power-speech-state-preemption-zh","川普 AI 框架瞄準電力、言論與州權","2026-03-26T07:12:18.695466+00:00",{"id":89,"slug":90,"title":91,"created_at":92},"ea6be18b-c903-4e54-97b7-5f7447a612e0","nvidia-gtc-2026-big-ai-announcements-zh","NVIDIA GTC 2026 重點拆解","2026-03-26T07:14:26.62638+00:00",{"id":94,"slug":95,"title":96,"created_at":97},"4bcec76f-4c36-4daa-909f-54cd702f7c93","claude-users-spreading-out-and-getting-better-zh","Claude 用戶更分散，也更會用","2026-03-26T07:22:52.325888+00:00",{"id":99,"slug":100,"title":101,"created_at":102},"bd903b15-2473-4178-9789-b7557816e535","openclaw-raises-hard-question-for-ai-models-zh","OpenClaw 逼問 AI 模型價值","2026-03-26T07:24:54.707486+00:00",{"id":104,"slug":105,"title":106,"created_at":107},"eeac6b9e-ad9d-4831-8eec-8bba3f9bca6a","gap-google-gemini-checkout-fashion-search-zh","Gap 把結帳搬進 Gemini","2026-03-26T07:28:23.937768+00:00",{"id":109,"slug":110,"title":111,"created_at":112},"0740e53f-605d-4d57-8601-c10beb126f3c","google-pushes-gemini-transition-to-march-2026-zh","Google 把 Gemini 轉換延到 2026 年 3…","2026-03-26T07:30:12.825269+00:00",{"id":114,"slug":115,"title":116,"created_at":117},"e660d801-2421-4529-8fa9-86b82b066990","metas-llama-4-benchmark-scandal-gets-worse-zh","Meta Llama 4 分數風波又擴大","2026-03-26T07:34:21.156421+00:00",{"id":119,"slug":120,"title":121,"created_at":122},"183f9e7c-e143-40bb-a6d5-67ba84a3a8bc","accenture-mistral-ai-sovereign-enterprise-deal-zh","Accenture 攜手 Mistral AI 賣主權 AI","2026-03-26T07:38:14.818906+00:00",{"id":124,"slug":125,"title":126,"created_at":127},"191d9b1b-768a-478c-978c-dd7431a38149","mistral-ai-faces-its-hardest-year-yet-zh","Mistral AI 迎來最硬的一年","2026-03-26T07:40:23.716374+00:00"]