[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-half-price-ai-real-frontier-smarter-models-zh":3,"article-related-half-price-ai-real-frontier-smarter-models-zh":31,"series-industry-8ffc6905-3e5a-4236-a031-bda41472e78d":79},{"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":23,"views":27,"created_at":28,"published_at":29,"topic_cluster_id":30},"8ffc6905-3e5a-4236-a031-bda41472e78d","half-price-ai-real-frontier-smarter-models-zh","半價 AI 才是主戰場，不是更聰明的模型","\u003Cp data-speakable=\"summary\">76% 更低的 \u003Ca href=\"\u002Ftag\u002Ftoken\">token\u003C\u002Fa> 價格，證明 AI 競爭已從智力轉向成本。\u003C\u002Fp>\u003Cp>AI 不再是比誰最聰明，而是比誰便宜到足以被代理系統全天候調用。\u003C\u002Fp>\u003Ch2>第一個論點\u003C\u002Fh2>\u003Cp>Meta 的 Muse Spark 1.1 把輸入 token 定在每百萬 $1.25、輸出 token 定在每百萬 $4.25；xAI 的 Grok 4.5 也壓到 $2 與 $6。這不是折扣促銷，而是把 frontier 模型直接拉進企業採購表。當一個模型不再只被人類偶爾問答，而是被 \u003Ca href=\"\u002Ftag\u002Fagent\">agent\u003C\u002Fa> 不停搜尋、重試、摘要、執行時，單次價格差 1 美元，最後會變成月帳單差數十萬美元。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783801974628-606c.png\" alt=\"半價 AI 才是主戰場，不是更聰明的模型\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>市場對這件事的反應已經很直接。BigGo 的來源顯示，平均 token 價格指數在一個月內從 5 月的每百萬 $2.10 降到 7 月 7 日的 $1.64，跌幅 22%。這種變化不是消費者在追新鮮感，而是買方開始用雲端基礎設施的方式看 AI。當採購\u003Ca href=\"\u002Fnews\u002F5-chatgpt-team-work-zh\">團隊\u003C\u002Fa>拿模型去和儲存、頻寬、GPU 使用量一起比較，\u003Ca href=\"\u002Ftag\u002Fbenchmark\">benchmark\u003C\u002Fa> 的光環就不再能單獨決定勝負。\u003C\u002Fp>\u003Ch2>第二個論點\u003C\u002Fh2>\u003Cp>agent 會把 token 成本變成董事會等級的問題。一般 chatbot 回答一次就結束；agent 卻會搜尋、草擬、修正、執行、再摘要，直到任務完成。Gartner 預測，企業軟體中使用 \u003Ca href=\"\u002Ftag\u002Fai-agents\">AI agents\u003C\u002Fa> 的比例，會從去年不到 5% 升到今年底的 40%。這代表 token 消耗不是線性成長，而是被工作流放大。對這種場景來說，便宜 60% 或 76% 不是省錢而已，而是決定功能能不能上線、能不能持續跑。\u003C\u002Fp>\u003Cp>Meta 內部的說法也很關鍵。Alexandr Wang 把 Muse Spark 1.1 稱為 “workhorse”，不是展示品。這句話等於承認：公司不是要先贏一場 demo 比賽，而是要拿下 coding、任務執行、長流程工作這些高頻場景。這些場景看重的是穩定、可預測、可擴張的單位經濟，而不是偶爾多幾分 benchmark 分數。只要便宜且夠用的層級被先佔住，開發者就會先把預設流量放進去。\u003C\u002Fp>\u003Ch2>第一個論點\u003C\u002Fh2>\u003Cp>價格戰之所以會把整個市場往下拉，是因為低價模型已經被訓練成可用。中國模型如 DeepSeek、Zhipu AI、MiniMax，早就把市場預期壓到比\u003Ca href=\"\u002Fnews\u002Fregulation-moves-bitcoin-ethereum-more-than-war-zh\">美國\u003C\u002Fa> frontier 產品低 60% 到 90%。OpenRouter 的資料顯示，自 2 月以來，中國模型拿下每週 token 使用量超過 30%。這代表開發者\u003Ca href=\"\u002Fnews\u002Fchina-winning-ai-cold-war-building-stack-zh\">不是在\u003C\u002Fa>投票支持某個品牌，而是在投票支持能把工作做完、而且不燒錢的模型。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783801966945-5157.png\" alt=\"半價 AI 才是主戰場，不是更聰明的模型\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>一旦低價模型足以處理摘要、客服、資料整理，昂貴的高階模型就會失去定價權。\u003Ca href=\"\u002Ftag\u002Fopenai\">OpenAI\u003C\u002Fa> 把 GPT-5.6 拆成三個層級，甚至推出每百萬輸入 $1、輸出 $6 的便宜方案，讀起來就是防守。Stanford AI Index 也指出，頂級模型之間的性能差距已經縮到很窄。當品質趨於收斂，價格就會變成唯一真正能拉開採用率的變數。\u003C\u002Fp>\u003Ch2>第二個論點\u003C\u002Fh2>\u003Cp>這場轉向不是理論，而是已經發生在使用量上。當模型被嵌進 agent、RAG、批次處理與內部工具鏈，企業看的是每完成一個任務要花多少錢，而不是某次測試答對了幾題。對 PM 來說，這意味著產品規格要從「我們用了哪個最強模型」改成「每個工作流的完成成本是多少」。對創辦人來說，這意味著毛利不再只受人力影響，也受 token 供應鏈影響。\u003C\u002Fp>\u003Cp>更重要的是，這種壓力會一路傳到基礎模型公司。只要客戶能把簡單任務切到低價模型，高價模型就只能保留給少數高風險任務，例如法律分析、進階程式推理、財務判讀。這不是高階模型失敗，而是市場成熟。真正的 frontier 不在於誰能把分數再推高一點，而在於誰能把足夠好的能力做成可大規模部署的價格。\u003C\u002Fp>\u003Ch2>反方可能怎麼說\u003C\u002Fh2>\u003Cp>最強的反對意見是：便宜不代表可用。企業在意的是可靠性、安全性與深度推理，高風險工作仍然需要最強模型。若便宜模型產生錯誤，後續的人工作業、合規風險與返工成本，會把省下來的錢全部吃掉。從這個角度看，價格戰可能只是把市場推向一堆快但淺的系統，反而拖慢真正的能力進步。\u003C\u002Fp>\u003Cp>這個批評成立，但它沒有推翻本文的結論，反而把答案說得更清楚：市場正在分層。高階模型會留在少數高風險任務，低價且夠用的模型會吃掉絕大多數重複性工作。對企業來說，真正的大宗成本不在少數高價推理，而在大量日常調用。只要這一層持續擴張，價格就會比單次智力表現更能決定誰拿到部署量。\u003C\u002Fp>\u003Ch2>你能做什麼\u003C\u002Fh2>\u003Cp>如果你是工程師、PM 或創辦人，別再把 AI 當成單一模型選型題，而要把它當成營運成本題。把簡單任務分流到低價模型，把高風險推理留給高階模型，並且用「完成一個任務的總成本」來衡量，而不是只看每次呼叫的驚喜感。這個市場最後贏的，不會是最會做 demo 的團隊，而是最會把 intelligence 當成可分配算力來管理的團隊。\u003C\u002Fp>","Meta 與 xAI 的定價戰證明，AI 競爭正在從 benchmark 轉向成本；能穩定降價、支撐代理工作流的模型，才會拿到真正的部署量。","finance.biggo.com","https:\u002F\u002Ffinance.biggo.com\u002Fnews\u002F93012dfd-f17e-4a3e-88f9-3973e47cdb76",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783801974628-606c.png","industry","zh","3e0651b7-5a37-4615-8462-5c695356154f",[17,18,19,20,21,22],"AI 定價","token 成本","agents","Meta","xAI","企業採用",[24,25,26],"AI 競爭重心已從 benchmark 轉向 token 單價與總成本。","agent 工作流會放大用量，讓價格成為採購與部署的核心指標。","市場正在分層：高階模型留給高風險任務，低價模型吃掉大宗工作。",0,"2026-07-11T20:32:23.661553+00:00","2026-07-11T20:32:23.652+00:00","f2c5fdb9-8e47-498a-ad3d-1e7ab235a0c4",{"tags":32,"relatedLang":38,"relatedPosts":42},[33,34,36],{"name":19,"slug":19},{"name":21,"slug":35},"xai",{"name":20,"slug":37},"meta",{"id":15,"slug":39,"title":40,"language":41},"half-price-ai-real-frontier-smarter-models-en","Half-price AI is the real frontier, not smarter models","en",[43,49,55,61,67,73],{"id":44,"slug":45,"title":46,"cover_image":47,"image_url":47,"created_at":48,"category":13},"d1753385-8c03-4dec-b939-e5ca8bae9030","opensearch-vector-search-benchmark-5-parts-zh","OpenSearch 向量搜尋基準的 5 種跑法","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783850566022-b79s.png","2026-07-12T10:02:22.269045+00:00",{"id":50,"slug":51,"title":52,"cover_image":53,"image_url":53,"created_at":54,"category":13},"6e790897-c9af-402c-a928-f2b0cc02f4e6","vector-databases-work-in-production-zh","4 種能上線的向量資料庫選擇","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783846963245-35py.png","2026-07-12T09:02:23.058273+00:00",{"id":56,"slug":57,"title":58,"cover_image":59,"image_url":59,"created_at":60,"category":13},"e5ae86b4-4434-48d4-86b4-146f609ce0a2","eu-ai-act-hits-business-systems-aug-2-2026-zh","歐盟 AI 法案上路前，企業先看這 5 件事","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783845168794-qyhi.png","2026-07-12T08:32:24.43396+00:00",{"id":62,"slug":63,"title":64,"cover_image":65,"image_url":65,"created_at":66,"category":13},"bc30f927-a6c9-4cdd-b734-6e8cd0b8265a","us-ai-law-2026-compliance-overview-zh","2026 美國 AI 法規控管地圖","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783843397232-2oow.png","2026-07-12T08:02:50.480302+00:00",{"id":68,"slug":69,"title":70,"cover_image":71,"image_url":71,"created_at":72,"category":13},"4d5e43ec-56bf-4ddf-aca0-e3b31065f132","webx-2026-agenda-stablecoins-ai-zh","WebX 2026 將穩定幣與 AI 推上主舞台","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783841563555-spp1.png","2026-07-12T07:32:24.035669+00:00",{"id":74,"slug":75,"title":76,"cover_image":77,"image_url":77,"created_at":78,"category":13},"4647fcd1-fee7-4819-958a-73a92587227a","gpt-56-full-suite-work-entry-openai-zh","GPT-5.6 全家桶不是炫技，是 OpenAI 的工作入口","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783837980636-jmwn.png","2026-07-12T06:32:32.520158+00:00",[80,85,90,95,100,105,110,115,120,125],{"id":81,"slug":82,"title":83,"created_at":84},"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":86,"slug":87,"title":88,"created_at":89},"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":91,"slug":92,"title":93,"created_at":94},"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":96,"slug":97,"title":98,"created_at":99},"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":101,"slug":102,"title":103,"created_at":104},"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":106,"slug":107,"title":108,"created_at":109},"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":111,"slug":112,"title":113,"created_at":114},"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":116,"slug":117,"title":118,"created_at":119},"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":121,"slug":122,"title":123,"created_at":124},"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":126,"slug":127,"title":128,"created_at":129},"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"]