[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-zilliz-cost-aware-scoring-vdbbench-zh":3,"article-related-zilliz-cost-aware-scoring-vdbbench-zh":33,"series-tools-dff07cae-e3a9-46e4-b069-c0ab3b7e18e3":78},{"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":26,"views":30,"created_at":31,"published_at":32,"topic_cluster_id":11},"dff07cae-e3a9-46e4-b069-c0ab3b7e18e3","zilliz-cost-aware-scoring-vdbbench-zh","VDBBench 把成本納入主指標","\u003Cp data-speakable=\"summary\">Zilliz 更新了開源 VDBBench，把向量資料庫評測從只看速度，改成同時看成本。\u003C\u002Fp>\u003Cp>\u003Ca href=\"https:\u002F\u002Fzilliz.com\" target=\"_blank\" rel=\"noopener\">Zilliz\u003C\u002Fa> 這次動的是評測邏輯，不是包裝詞。\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fzilliztech\u002FVectorDBBench\" target=\"_blank\" rel=\"noopener\">VectorDBBench\u003C\u002Fa> 原本偏重速度與吞吐，現在把成本拉進同一張表。對做 \u003Ca href=\"\u002Ftag\u002Frag\">RAG\u003C\u002Fa>、語意搜尋、推薦系統的人來說，這很實際，因為雲端帳單才是最後會被財務盯上的數字。\u003C\u002Fp>\u003Cp>這次更新也很符合現在的採購現場。很多團隊在 demo 裡看到漂亮的延遲數字，到了正式上線才發現記憶體、複本、儲存和維運人力一起把費用拉高。VDBBench 把成本放進評分，等於逼大家正視「跑得快」和「跑得貴」之間的差距。\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>指標\u003C\u002Fth>\u003Cth>更新內容\u003C\u002Fth>\u003Cth>實際意義\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>評測重點\u003C\u002Ftd>\u003Ctd>成本納入核心評分\u003C\u002Ftd>\u003Ctd>可直接比每美元效能\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>專案性質\u003C\u002Ftd>\u003Ctd>開源、供應商中立\u003C\u002Ftd>\u003Ctd>降低單一廠商的話術偏差\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>主要場景\u003C\u002Ftd>\u003Ctd>向量資料庫評測\u003C\u002Ftd>\u003Ctd>對 RAG 與語意搜尋很有用\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>決策用途\u003C\u002Ftd>\u003Ctd>從技術比較走向採購判斷\u003C\u002Ftd>\u003Ctd>更貼近真實上線需求\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>為什麼向量資料庫不能只看速度\u003C\u002Fh2>\u003Cp>向量資料庫在 AI 堆疊裡的位置很尷尬。它常常不是最吸睛的那層，卻是最容易燒錢的那層。你可以在小測試裡看到很漂亮的 QPS，但一旦資料量放大、查詢並發升高、還要維持高可用，成本就會開始翻臉。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786651373076-fihi.png\" alt=\"VDBBench 把成本納入主指標\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>這也是成本評分有價值的地方。它讓團隊不只問「快不快」，而是直接問「這個速度值不值這個錢」。如果答案不漂亮，架構會議就會少掉很多自我感動。\u003C\u002Fp>\u003Cp>對產品團隊來說，這件事更直接。你要的是可上線的方案，不是簡報上好看的圖。當 \u003Ca href=\"\u002Ftag\u002Fbenchmark\">benchmark\u003C\u002Fa> 開始把成本算進去，選型討論就會更接近真實世界。\u003C\u002Fp>\u003Cul>\u003Cli>低延遲不代表低總成本。\u003C\u002Fli>\u003Cli>高吞吐不代表適合長期營運。\u003C\u002Fli>\u003Cli>成本評分能把雲端、記憶體和複本差異攤開。\u003C\u002Fli>\u003Cli>同一組資料，跑法不同，帳單也會差很多。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>VDBBench 想修正什麼問題\u003C\u002Fh2>\u003Cp>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fzilliztech\u002FVectorDBBench\" target=\"_blank\" rel=\"noopener\">VDBBench\u003C\u002Fa> 的定位很清楚，就是做向量資料庫的基準測試，而且盡量保持中立。這點很重要，因為 benchmark 一旦被廠商話術帶著走，結果就只剩宣傳稿\u003Ca href=\"\u002Fnews\u002Fpixel-11-launch-highlights-gemini-features-zh\">功能\u003C\u002Fa>。開源工具至少讓大家能看方法、改參數、重跑測試。\u003C\u002Fp>\u003Cp>把成本放進來之後，VDBBench 的用途也變了。以前它比較像排行榜，現在更像選型工具。平台團隊可以拿它來比預算，應用團隊可以拿它來看延遲與召回是否值得那個價錢。\u003C\u002Fp>\u003Cp>這種轉向其實很符合 \u003Ca href=\"\u002Ftag\u002Fai-\">AI 基礎設施\u003C\u002Fa>的現況。大家早就不太相信只講吞吐、不講花費的圖表。真正要簽約時，大家看的都是每月花多少、擴容難不難、維運會不會把工程師拖垮。\u003C\u002Fp>\u003Cblockquote>“Benchmarks should measure what users actually pay for, not just what vendors want to show,” said Zilliz founder and CEO Charles Xie.\u003C\u002Fblockquote>\u003Cp>這句話很直白，也很到位。評測如果只量速度，常常會把昂貴方案包裝得很好看。把成本拉進來後，很多看起來漂亮的數字就沒那麼神了。\u003C\u002Fp>\u003Ch2>這會怎麼改變供應商比較\u003C\u002Fh2>\u003Cp>向量資料庫的比較，從來都不是單看一個數字。兩套系統可能延遲差不多，召回也差不多，但其中一套需要更多 RAM、更多副本，或更貴的雲端配置。只看 throughput，這些差異很容易被藏起來。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786651377376-b3pr.png\" alt=\"VDBBench 把成本納入主指標\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>成本評分的作用，就是把這些差異攤平。這對採購很重要，因為最後拿到預算的人，不會因為你的 benchmark 圖很好看就多給錢。對工程團隊來說，也能少踩一些「測得過、養不起」的坑。\u003C\u002Fp>\u003Cp>如果你現在在比較方案，下面這幾個名字大概都會出現在名單上。每個產品的部署方式、定價邏輯、維運成本都不一樣，拿同一把尺量才有意義。\u003C\u002Fp>\u003Cul>\u003Cli>\u003Ca href=\"https:\u002F\u002Fmilvus.io\" target=\"_blank\" rel=\"noopener\">Milvus\u003C\u002Fa>：Zilliz 自家的向量資料庫，常被拿來和其他引擎一起比較。\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fwww.pinecone.io\" target=\"_blank\" rel=\"noopener\">Pinecone\u003C\u002Fa>：託管型方案，很多團隊會先拿它當基準。\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fweaviate.io\" target=\"_blank\" rel=\"noopener\">Weaviate\u003C\u002Fa>：語意搜尋場景常見的比較對象。\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Ffaiss\" target=\"_blank\" rel=\"noopener\">FAISS\u003C\u002Fa>：研究和自建系統常用，彈性高，但整合成本也要算。\u003C\u002Fli>\u003C\u002Ful>\u003Cp>我覺得這次更新最有意思的地方，在於它把 benchmark 從「誰最快」拉回「誰最划算」。這種變化看起來不華麗，卻很接近真實世界。畢竟 AI 系統不是跑一次 demo 就結束，是真金白銀每天在花。\u003C\u002Fp>\u003Ch2>產業裡為什麼開始重視每美元效能\u003C\u002Fh2>\u003Cp>這幾年 AI 基礎設施的問題很一致。模型越來越大，資料越來越多，查詢越來越密，成本也越來越難壓。團隊一開始可能只想把功能做出來，後來就會發現，真正難的是把服務穩穩養住。\u003C\u002Fp>\u003Cp>所以現在很多基礎設施工具都在往「真實工作負載」靠。大家不太想再看只適合實驗室的數字，因為那種數字對上線幫助有限。VDBBench 加入成本評分，正好踩在這個方向上。\u003C\u002Fp>\u003Cp>這也提醒一件事：AI 軟體選型不能只看技術白皮書。你要看資料量、查詢型態、延遲目標、部署環境，還要看團隊有沒有能力\u003Ca href=\"\u002Fnews\u002Fanthropic-custom-ai-inference-chips-claude-zh\">自己\u003C\u002Fa>維運。少掉任何一項，成本都可能失真。\u003C\u002Fp>\u003Ch2>接下來該怎麼用這個更新\u003C\u002Fh2>\u003Cp>如果你現在正在評估向量資料庫，我會\u003Ca href=\"\u002Fnews\u002Faws-continuum-turns-ai-coding-into-safer-fixes-zh\">建議\u003C\u002Fa>直接改測法。把 latency、recall、throughput 和成本放在同一個 workload 裡看。不要用 1 萬筆資料的測試去推 1 億筆資料的結論，那樣很容易自欺欺人。\u003C\u002Fp>\u003Cp>更實際的做法，是先定義你的上線條件，再回頭跑 benchmark。你的查詢峰值是多少，容忍延遲是多少，資料更新頻率多高，這些都要先寫清楚。只要條件不同，排名就可能整個翻盤。\u003C\u002Fp>\u003Cp>我會把這次更新看成一個提醒。AI 基礎設施的評測，正在從炫技走向算帳。下一次你看到某個資料庫宣稱自己最快，先問一句：每月帳單是多少。這個問題通常比跑分更有用。\u003C\u002Fp>","Zilliz 更新開源 VDBBench，把向量資料庫評測從只看速度，改成同時看成本。這讓團隊能用每美元效能比較不同方案，也更貼近實際採購與上線需求。","martechseries.com","https:\u002F\u002Fmartechseries.com\u002Fanalytics\u002Fdata-management-platforms\u002Fzilliz-adds-cost-aware-benchmarking-to-vdbbench-the-open-source-vector-database-benchmark\u002F",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786651373076-fihi.png","tools","zh","04620d43-b903-445e-bc49-7bb6b9f6c019",[17,18,19,20,21,22,23,24,25],"VDBBench","Zilliz","向量資料庫","成本評分","benchmark","Milvus","Pinecone","Weaviate","FAISS",[27,28,29],"VDBBench 把成本納入核心評分，讓向量資料庫能比每美元效能。","這種評測方式更貼近 RAG、語意搜尋和正式上線的採購需求。","團隊選型時，應同時看 latency、recall、throughput 和總成本。",0,"2026-08-13T20:02:30.752194+00:00","2026-08-13T20:02:30.74+00:00",{"tags":34,"relatedLang":37,"relatedPosts":41},[35,36],{"name":21,"slug":21},{"name":19,"slug":19},{"id":15,"slug":38,"title":39,"language":40},"zilliz-cost-aware-scoring-vdbbench-en","Zilliz adds cost-aware scoring to VDBBench","en",[42,48,54,60,66,72],{"id":43,"slug":44,"title":45,"cover_image":46,"image_url":46,"created_at":47,"category":13},"dda641de-50ed-419d-81dc-d8f8edc33084","vdbbench-cost-benchmark-vector-databases-zh","VDBBench 把向量資料庫比法改掉","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786654993634-4al2.png","2026-08-13T21:02:47.275149+00:00",{"id":49,"slug":50,"title":51,"cover_image":52,"image_url":52,"created_at":53,"category":13},"c55af4bc-dfce-4031-896d-89c6e76ba2c0","zilliz-cost-aware-vdbbench-benchmark-zh","Zilliz替VDBBench加入成本指標","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786653169746-xfnk.png","2026-08-13T20:32:26.208471+00:00",{"id":55,"slug":56,"title":57,"cover_image":58,"image_url":58,"created_at":59,"category":13},"acd0364e-15cb-4a7c-b5d4-e670436ed521","pixel-11-launch-highlights-gemini-features-zh","Pixel 11 發表重點與 Gemini 新功能","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786649574526-g2vc.png","2026-08-13T19:32:23.509251+00:00",{"id":61,"slug":62,"title":63,"cover_image":64,"image_url":64,"created_at":65,"category":13},"cad5997c-d40d-4fac-9b39-e1f86a326107","aws-continuum-turns-ai-coding-into-safer-fixes-zh","AWS Continuum 把修漏洞變安全建議","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786644251974-emsc.png","2026-08-13T18:03:30.983335+00:00",{"id":67,"slug":68,"title":69,"cover_image":70,"image_url":70,"created_at":71,"category":13},"9cbd8e8a-df48-4e25-a950-2a8552a11c1e","open-generative-ai-github-studio-breakdown-zh","Open-Generative-AI 讓 GitHub 變工作室","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786622603725-o8ad.png","2026-08-13T12:02:47.120311+00:00",{"id":73,"slug":74,"title":75,"cover_image":76,"image_url":76,"created_at":77,"category":13},"98e7c6dd-38f6-4740-bb79-20985bff3f9a","benchmark-scores-dont-predict-your-bill-zh","Benchmark 分數不等於帳單","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786496578340-z656.png","2026-08-12T01:02:41.398372+00:00",[79,84,89,94,99,104,109,114,119,124],{"id":80,"slug":81,"title":82,"created_at":83},"855cd52f-6fab-46cc-a7c1-42195e8a0de4","surepath-real-time-mcp-policy-controls-zh","SurePath 推出即時 MCP 政策控管","2026-03-26T07:57:40.77233+00:00",{"id":85,"slug":86,"title":87,"created_at":88},"9b19ab54-edef-4dbd-9ce4-a51e4bae4ebb","mcp-in-2026-the-ai-tool-layer-teams-use-zh","2026 年 MCP：團隊真的在用的 AI 工具層","2026-03-26T08:01:46.589694+00:00",{"id":90,"slug":91,"title":92,"created_at":93},"af9c46c3-7a28-410b-9f04-32b3de30a68c","prompting-in-2026-what-actually-works-zh","2026 提示工程，真正有用的是什麼","2026-03-26T08:08:12.453028+00:00",{"id":95,"slug":96,"title":97,"created_at":98},"05553086-6ed0-4758-81fd-6cab24b575e0","garry-tan-open-sources-claude-code-toolkit-zh","Garry Tan 開源 Claude Code 工具包","2026-03-26T08:26:20.068737+00:00",{"id":100,"slug":101,"title":102,"created_at":103},"042a73a2-18a2-433d-9e8f-9802b9559aac","github-ai-projects-to-watch-in-2026-zh","2026 必看 20 個 GitHub AI 專案","2026-03-26T08:28:09.619964+00:00",{"id":105,"slug":106,"title":107,"created_at":108},"a5f94120-ac0d-4483-9a8b-63590071ac6a","claude-code-vs-cursor-2026-zh","Claude Code 與 Cursor 深度對比：202…","2026-03-26T13:27:14.279193+00:00",{"id":110,"slug":111,"title":112,"created_at":113},"0975afa1-e0c7-4130-a20d-d890eaed995e","practical-github-guide-learning-ml-2026-zh","2026 機器學習入門 GitHub 實用指南","2026-03-27T01:16:49.712576+00:00",{"id":115,"slug":116,"title":117,"created_at":118},"bfdb467a-290f-4a80-b3a9-6f081afb6dff","aiml-2026-student-ai-ml-lab-repo-review-zh","AIML-2026：像課綱的學生實驗 Repo","2026-03-27T01:21:51.467798+00:00",{"id":120,"slug":121,"title":122,"created_at":123},"80cabc3e-09fc-4ff5-8f07-b8d68f5ae545","ai-trending-github-repos-and-research-feeds-zh","AI Trending：把 AI 資源收成一張表","2026-03-27T01:31:35.262183+00:00",{"id":125,"slug":126,"title":127,"created_at":128},"3ce6e6e2-bac5-463e-9f8d-45caabcc61f7","awesome-ai-for-science-research-tools-map-zh","AI 科研工具清單，開始像地圖了","2026-03-27T01:46:50.521945+00:00"]