[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-why-kimi-k26-changes-coding-model-race-zh":3,"article-related-why-kimi-k26-changes-coding-model-race-zh":36,"series-model-release-da646ee1-db2e-48b4-9ffe-b79a8a81ae93":87},{"id":4,"title":5,"content":6,"summary":7,"source":8,"source_url":9,"author":10,"image_url":11,"keywords":12,"language":19,"translated_content":10,"views":20,"is_premium":21,"created_at":22,"updated_at":22,"cover_image":11,"published_at":23,"rewrite_status":24,"rewrite_error":10,"rewritten_from_id":25,"slug":26,"category":27,"related_article_id":28,"status":29,"google_indexed_at":10,"x_posted_at":10,"tweet_text":10,"title_rewritten_at":10,"title_original":10,"key_takeaways":30,"topic_cluster_id":34,"embedding":35,"is_canonical_seed":21},"da646ee1-db2e-48b4-9ffe-b79a8a81ae93","為什麼 Kimi K2.6 會改寫寫程式模型競賽","\u003Cp data-speakable=\"summary\">Kimi K2.6 以開放權重和更低成本，逼近 GPT-5.5 的程式能力。\u003C\u002Fp>\u003Cp>Kimi K2.6 不是又一個「差不多」的模型，而是第一個把閉源前沿模型的定價邏輯打到失血的開放權重編碼模型。\u003Ca href=\"\u002Ftag\u002Fmoonshot-ai\">Moonshot AI\u003C\u002Fa> 在 2026 年 4 月 20 日發布它，重點很直接：在 \u003Ca href=\"\u002Ftag\u002Fswe-bench\">SWE-Bench\u003C\u002Fa> Pro 上拿到 58.6，追平 GPT-5.5，同時 API 價格約為每百萬 input token 0.95 美元、output token 4 美元；對照 GPT-5.5 的 5 美元與 30 美元，這不是小折扣，而是採購、產品設計與團隊默認模型選擇的改寫。\u003C\u002Fp>\u003Ch2>第一個論點：它打中的就是工程師最痛的地方\u003C\u002Fh2>\u003Cp>真正有價值的不是玩具題，而是能不能修真實程式碼庫。SWE-Bench Pro 測的是模型是否能在真實 repo 裡修 issue，K2.6 在這項指標上追平 GPT-5.5，這件事的意義很大。因為工程團隊買的不是聊天能力，而是能不能把 patch 送進去、測試跑過、repo 不被搞亂。Moonshot 也公布 K2.6 相較 K2.5 的長程編碼完成能力提升 185%，Terminal-Bench 2.0 從 50.8 拉到 66.7，顯示它不是只會短答，而是針對持續工作優化。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778980423319-9xts.png\" alt=\"為什麼 Kimi K2.6 會改寫寫程式模型競賽\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>產品形態也說明了方向。K2.6 不是單純 autocomplete，而是為 age\u003Ca href=\"\u002Fnews\u002Fanthropic-small-business-push-threat-to-saas-zh\">nt\u003C\u002Fa>ic coding 設計：支援 vision input、上下文長度到 256K tokens，能把一個任務維持在長時間工作脈絡裡。對實務團隊來說，這意味著它更接近「丟給它一個 refactor，過一段時間回來看結果」的工作模式，而不是只能做局部補字。當你的日常是 bug fix、migration、integration，這種差異不是錦上添花，而是工作流本身的改變。\u003C\u002Fp>\u003Ch2>第二個論點：價格差已經大到不能裝沒看見\u003C\u002Fh2>\u003Cp>成本是這場競賽真正的分水嶺。以 Moonshot 自家 API 計價來看，K2.6 的 input 成本大約是 GPT-5.5 的五分之一，output 成本約是七分之一。Cla\u003Ca href=\"\u002Fnews\u002Fru-he-yong-cuda-oxide-jian-li-rust-gpu-he-xin-zh\">ud\u003C\u002Fa>e Opus 4.7 雖然在 SWE-Bench Pro 上更高，達到 64.3，但它的價格也更高，約每百萬 input tokens 5 美元、output tokens 25 美元。若你的團隊每天都在產生程式碼，模型選擇就不再是品牌偏好，而是 burn r\u003Ca href=\"\u002Fnews\u002Fanthropic-gates-foundation-200m-deal-zh\">ate\u003C\u002Fa> 問題。\u003C\u002Fp>\u003Cp>開放權重把經濟性再往前推一步。K2.6 已上架 \u003Ca href=\"\u002Ftag\u002Fhugging-face\">Hugging Face\u003C\u002Fa>，採 modified MIT license，代表團隊可以自架、微調，並在多數情境下避免 vendor lock-in。Moonshot 只對月活超過 1 億或月營收超過 2000 萬美元的產品加上 UI branding 要求；對大多數公司來說，它的使用條件接近真正的開放資產。當你可以把模型放進自己的基礎設施，最便宜的 token 就是根本不用送去閉源 API 的 token。\u003C\u002Fp>\u003Ch2>第三個論點：它把 agent 工作從框架層拉回模型層\u003C\u002Fh2>\u003Cp>K2.6 最重要的不是分數，而是 300-agent swarm 系統。Moonshot 的說法是，它能把任務拆給數百個 sub-agents，協調最多 4,000 個步驟，單次自主 session 可跑 13 小時。這等於直接挑戰一個常見假設：嚴肅的 agent 工作必須依賴昂貴閉源模型加上外部 orchestration。K2.6 把平行化變成模型能力，而不是工程團隊額外要付的框架稅。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778980429794-j6w2.png\" alt=\"為什麼 Kimi K2.6 會改寫寫程式模型競賽\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>這件事之所以重要，是因為 orchestration 正是很多 \u003Ca href=\"\u002Ftag\u002Fai-coding\">AI coding\u003C\u002Fa> 專案失敗的地方。多數團隊不是缺一個會講話的 chatbot，而是缺一個能拆任務、分派子工作、恢復中斷、維持狀態的系統。K2.6 把這些能力往模型內收，代表產品團隊少寫一層脆弱 glue code，工程師少維護一堆容易壞的 agent pipeline，創辦人則多了一個能同時壓低建置時間與 token 支出的選項。\u003C\u002Fp>\u003Ch2>反方可能怎麼說\u003C\u002Fh2>\u003Cp>最強的反對意見其實很合理：K2.6 不是整體最強的模型。Claude Opus 4.7 在 SWE-Bench Pro 上領先，GPT-5.5 在更廣泛的 intelligence 指標上也更強，而且兩者都有更大的 context window。若你在做高風險系統，絕對品質比成本更重要，那麼付費買最強閉源模型是理性的選擇。再者，開放權重不會自動消除評測、guardrails 與人工審核的需求，agent swarm 也可能把錯誤放大得跟效率一樣快。\u003C\u002Fp>\u003Cp>這些批評成立，但不足以推翻 K2.6 的價值。問題從來不是它是否在每一個維度都擊敗所有前沿模型，它沒有；問題是開放模型是否已經能以遠低於閉源前沿模型的成本，完成第一梯隊的程式工作。這一點上，K2.6 已經贏了。它在最貼近工程師痛點的 \u003Ca href=\"\u002Ftag\u002Fbenchmark\">benchmark\u003C\u002Fa> 上追平 GPT-5.5，又同時提供開放權重與激進定價，足以成為大量 coding workload 的預設選項。\u003C\u002Fp>\u003Ch2>你能做什麼\u003C\u002Fh2>\u003Cp>如果你是工程師，這週就拿 K2.6 跑一個真實 repo 任務：修一個 bug、做一次 migration，或補一個需要測試與後續修改的 feature。如果你是 PM，把它和現有 assistant 在一個需要多輪完成的任務上比成本與完成率。如果你是創辦人，把 K2.6 當成產品槓桿，不要只當新奇玩具。它最適合放在長程編碼、自主研究、多步驟 agent 工作這些能同時降工時與降 token 花費的地方。市場已經多了一個便宜但夠硬的選項，現在忽視它，就是預算判斷失誤。\u003C\u002Fp>","Kimi K2.6 以開放權重、接近 GPT-5.5 的程式能力和更低成本，逼迫編碼模型競賽重新定價。","felloai.com","https:\u002F\u002Ffelloai.com\u002Fkimi-k2-6-is-here-the-open-source-ai-model-tying-gpt-5-5-on-coding\u002F",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778980423319-9xts.png",[13,14,15,16,17,18],"Kimi K2.6","GPT-5.5","SWE-Bench Pro","開放權重","程式編碼模型","agentic coding","zh",1,false,"2026-05-17T01:13:21.45526+00:00","2026-05-17T01:13:21.434+00:00","done","900f809a-196b-41b3-a0bf-a17bde5a6cef","why-kimi-k26-changes-coding-model-race-zh","model-release","792ebbf9-6c9b-4651-9a00-c45ad71248fd","published",[31,32,33],"Kimi K2.6 以開放權重和低價格，逼近 GPT-5.5 的程式能力。","它在真實程式修復、長程編碼與 agent 工作上，對工程團隊更有實用價值。","對多數團隊來說，K2.6 已經足以成為預設編碼模型，而不是備胎。","0ccb5d2e-69f1-4354-a3e0-cb370221cd95","[-0.008693948,0.009295687,0.0015332042,-0.09317331,-0.00702015,-0.0011094984,-0.0126239415,0.0073089963,0.0072979336,0.00678038,0.0030632494,-0.0015616746,0.025027206,-0.021904431,0.14549619,0.032260578,-0.010781569,0.014763446,-0.0006765899,0.013711919,0.0071435687,0.028325666,0.024611197,-0.012126951,-0.0013642954,-0.01667539,0.014533012,-0.014673511,0.029137686,0.027858507,-0.018402675,-0.022459416,0.0033672743,0.0055056345,-0.0017751001,0.0076043447,0.04438782,-0.005350227,0.0016964263,0.0013645191,0.0035506787,-0.016784137,0.025472369,-0.023619097,-0.01634104,-0.0014206044,-0.0049015917,-0.03690638,-0.016904792,0.011806627,-0.009063738,0.030585505,-0.013437935,-0.13937756,-0.012571893,0.04270897,-0.0006905778,0.0064322175,-0.011459316,0.0063522523,-0.006778015,0.0056266882,-0.016258713,0.003350369,0.0011468334,-0.018967226,0.013925447,0.0011509291,-0.016413713,-0.007796463,-0.039200854,-0.0016066679,0.0004252915,-0.00074670394,0.0095767155,-0.0423149,0.0019513072,-0.0007344561,0.010216015,-0.0050381874,0.027463505,0.0018378152,0.0058055143,-0.005615506,0.007701208,-0.007553615,-0.008434638,-0.030320385,0.015354109,-0.02049335,-0.011700343,0.010460369,0.0006547856,0.015498115,0.025676807,-0.004210573,-0.017135927,0.003482603,0.008128137,-0.003043166,0.00033204854,-0.022931479,0.013795663,0.0063887048,0.00012624811,0.0030180789,0.008077689,0.015015291,0.0065530566,0.018574465,-0.021744816,-0.0100293625,-0.025919665,0.0057657748,-0.003049486,-0.14787549,0.02459198,0.01196266,0.0027620143,-0.008642925,0.010514765,0.0007430664,0.008939255,0.03997962,0.015669042,0.014698731,-0.009531652,0.009052019,-0.004100191,0.034466673,-0.02520588,0.015293263,0.0019075832,-0.020637482,-0.004724373,0.0180161,0.01994206,-0.009082959,-0.020664586,-0.02349886,0.0006006952,0.018763961,-0.02610897,-0.0040375236,-0.018697308,-0.003731168,-0.018205168,0.002967956,0.0026662857,-0.013988438,0.017951556,-0.007706345,-0.0060501355,-0.01796339,-0.00047112335,-0.03318805,0.010067982,0.0053279786,-0.00013303189,0.021929799,-0.016004171,-0.012004866,-0.013968913,-0.0031717324,-0.0112461755,0.01721552,-0.022958253,0.022280747,0.005060728,0.005057622,0.00312556,0.014716291,-0.008724811,0.017690293,0.003884242,-0.022573616,0.0006935923,-0.015608239,0.00779986,-0.018408705,0.026752802,-0.0069743996,0.0003884536,-0.0031114263,-0.01950081,-0.023031304,0.003889263,-0.02089303,-0.02384079,-0.01830523,-0.021536753,-0.0065511954,0.022547055,-0.02653674,0.009449304,-0.011344409,-0.015608057,0.016157847,-0.010521569,0.019315548,0.016035894,-0.020086074,0.0093584815,-0.02034851,0.022842063,-0.023098554,0.0083667105,-0.005549524,-0.0002696993,-0.036471143,-0.027289925,0.025937079,0.027979685,-0.010915198,-0.02209544,0.015775952,-0.022486972,-0.04172023,0.005558042,0.0036395283,0.01690533,-0.03137567,0.006255635,0.009988686,-0.011022329,-0.03870387,-0.005719504,-0.004581383,-0.028784258,0.037293795,0.0237296,0.0006701732,0.016254952,-0.0074277734,0.018874375,-0.0196601,0.025708348,0.017060224,0.0017131839,0.027125467,-0.020196645,0.026615392,0.0071405754,7.577094e-05,0.016145296,-0.028576279,0.011623709,-0.011059097,-0.023085132,0.02050462,-0.012196933,-0.00017383453,0.0002212758,-0.011425327,0.02867002,-0.022127094,-0.00846693,-0.0155479,-0.03580402,0.008456364,-0.028629677,-0.008310379,0.0013001292,-0.012399557,-0.0014740084,0.009214386,0.028350726,-0.018647093,-0.01132731,-0.007399886,-0.026473423,0.020745952,0.006705102,-0.016270544,0.015526902,-0.01560401,-0.04557103,0.059545543,-0.010024307,0.0046395077,0.015471579,0.020210054,-0.01211387,0.009420537,-0.0067043635,-0.011826411,-0.018995669,0.011640261,0.0059343125,-0.012192126,0.026149733,0.02768354,0.0029840197,0.007965668,-0.009566322,0.00027725246,-0.008014169,0.013523646,0.005383197,0.017664107,0.0112463385,-0.0045775617,0.02561669,0.057878833,-0.0205683,0.003902445,0.012701654,-0.008818942,-0.02354078,-0.016897736,-0.012448159,-0.030287443,0.0028522608,-0.033725917,0.0039977906,-0.012175165,0.014865592,-0.023152329,-0.010041428,0.008449201,-0.005969436,-0.033988282,-6.064542e-06,-0.009837418,-0.012684048,0.020546027,0.010618243,0.0019501833,0.025400817,0.0077769803,0.0065930677,0.018011365,0.019427145,-0.00050739973,0.0025711646,-0.010337526,-0.011459111,0.011623395,-0.021243619,-0.021378577,0.004577854,0.005736566,-0.023885297,-0.008223251,-0.021549024,0.003966073,0.015583131,-0.00902621,-0.003160682,-0.007312506,0.005993928,-0.020800835,0.008268573,0.008209789,-0.036116116,0.014501231,0.000769231,0.0038445184,0.012894882,-0.023280475,0.042345434,-0.021405559,0.01465636,-0.027400408,0.025404185,0.0074289707,0.011001438,-0.0031943126,0.0066578924,-0.034631904,-0.0061428854,0.0054347,-0.0014182606,-0.019750226,0.010961387,-0.01118726,-0.016218215,0.001031062,0.026906613,-0.005085242,-0.018674359,0.042682126,-0.027362864,-0.011061632,0.02429745,0.001982932,0.027354648,0.014263172,0.0046878853,0.034948345,0.024218021,-0.013008686,0.023230223,-0.002375297,-0.020013783,-0.02101089,-0.026600417,-0.010583656,-0.01716782,-0.027218604,-0.0056121033,-0.007608019,-0.00621992,0.0097189015,-0.0074118185,0.005431216,0.00056435715,0.025172418,-0.008616793,0.016937064,-0.008900889,0.006388282,0.0091397045,-0.0015900086,0.00014878545,0.0062812725,0.009836982,0.016506322,0.012610996,0.019704986,-0.014996689,0.017671423,0.009856645,0.013461419,-0.0009395107,0.012544488,-0.022439312,0.0130135445,0.0025870237,-0.014088524,0.008662185,-0.007695718,-0.003845171,-0.013019338,-0.0009076492,-0.039791342,-0.027875002,0.016851151,-0.020883745,-0.009842967,-0.006669534,0.007297861,0.02420441,0.019762082,0.0021462638,0.007962191,0.0009344709,-0.019520128,-0.028319877,0.009992265,0.041528128,0.00019916812,0.059638888,-0.007971165,0.00021823737,0.020578664,0.006425847,0.007062747,-0.0017335328,0.005152313,-0.01867209,5.0781946e-05,0.020033574,0.028075974,0.0073488816,-0.019173516,-0.013503298,-0.013137111,-0.013206837,0.026483087,-0.010825887,-0.006635168,0.03311341,0.0019245956,-0.033051427,-0.016711824,0.021650184,-0.008113569,0.0109173665,0.0045892433,-0.005227843,-0.022735894,0.017488742,-0.0011800888,0.017610898,0.023739263,-0.0192937,-0.01727277,-0.013334559,-0.008029419,0.059427995,0.02346342,0.0058006044,0.0014761841,-0.02002373,-0.027526597,-0.0006133904,-0.052368674,0.01978819,0.0092482045,0.005977841,-0.021409985,0.021627618,-0.018046532,-0.008327244,-0.0010271556,0.010603637,-0.025984466,0.001820567,0.00060020725,-0.024124587,-0.0010385368,0.019936213,0.008815923,0.00037075335,0.012386531,-0.008067495,-0.008604654,-0.025568353,0.00208909,-0.013572647,0.009200893,0.0033827932,0.012127942,-0.0057928567,0.0056749443,0.033893324,0.018588986,0.011515025,0.013270351,0.00048743223,0.026761362,1.1265049e-05,-0.029298326,-0.008687328,0.0037625257,0.028296614,0.042170033,-0.018644998,0.0021092368,0.014362659,-0.025720788,-0.024267871,0.0010219762,-0.0033018116,-0.094618164,0.017095108,-0.011318147,-0.023450142,-0.0013292226,-0.014430226,0.012845817,-0.04639112,0.004150628,0.0126947295,0.030006597,0.007198694,-0.002662287,-0.001203166,-0.0044172877,-0.010578706,-0.011864749,-0.02678864,0.020550244,0.0006500935,0.013438846,0.0041110045,0.024341574,-0.0062157004,-0.017878244,-0.02711057,0.0057652057,-0.01755432,0.0028379413,-0.018312812,-0.046906758,-0.001669892,-0.0006767188,0.026447173,0.016738499,-0.00231067,-0.0032741036,-0.010008457,0.0160992,0.008105038,0.034277756,0.01803339,-0.0012784305,0.0035638148,0.005856224,-0.005565647,0.00048441184,-0.005591188,-0.0060323603,-0.009923482,-0.024138764,-0.007297985,0.004589589,-0.018957602,-0.0012879642,-0.009236282,0.008214331,-0.0073791384,0.011707311,0.007018562,-0.018462393,-0.008258005,-0.0097862305,0.016997775,-0.036238015,0.013264956,0.013632237,0.022998583,0.0053480756,-0.015496961,0.017717024,-0.010438736,-0.008829979,-0.001816325,0.012956838,-0.008941378,0.009255214,0.013528994,0.023810849,0.007543439,-0.031787112,-0.014912008,-0.09317227,-0.028403759,-0.010888026,-0.0012745081,-0.023233714,-0.016159765,-0.013002599,-0.030949306,0.0069437944,-0.01080629,0.010834917,-0.0021087646,0.0060774214,-0.039566208,0.021110795,-0.0024920732,-0.0035823274,0.015734639,0.008016533,-0.018800123,-0.012825065,0.023200613,0.025644597,-0.027956082,-0.013961504,0.028941777,0.004276398,0.018717239,0.011495819,0.0056247422,-0.0053984774,-0.12793283,-0.04027055,-0.0032876614,-0.015126051,0.0050284835,0.020344349,0.015392959,-0.0067351474,-0.016257582,-0.015330769,-0.0002445331,-0.02066425,0.00034796444,-0.009975442,0.022600576,0.09859433,0.015124272,-0.000791207,-0.00020155897,0.012398713,0.013058199,-0.005770634,-0.0056672273,0.010518334,0.012300256,-0.0033363858,0.019143188,0.0025504958,7.246212e-05,0.012592257,0.013767326,0.0009798832,0.0012401362,-0.021483833,0.0045076236,0.009535302,-0.03784016,-0.034387313,0.033304527,0.0075014676,0.007160007,0.015969375,0.012191246,0.015468192,-0.021766525,0.010571084,-0.010032315,-0.021237696,-0.009229723,0.004326649,-0.009270384,-0.06404609,-0.004555808,-0.009202128,0.023865683,-0.0022112718,0.017083023,0.013889402,0.015394079,-0.0012501831,0.00610578,0.00091646076,-0.0034842815,0.023370499,-0.029454146,0.037596177,0.018718928,0.005227691,0.0026025095,-0.0006054548,-0.008400393,0.010385288,0.004002585,0.0034342618,0.03223592,-0.039463036,0.0038968676,0.012101529,0.02863276,-0.021405965,0.018824888,0.0025264267,-0.018678587,0.001832401,0.015834482,0.017981136,0.035182945,-0.023002958,-0.00095597026,0.0049033994,-0.0012512489,0.008841593,-0.012798267,0.01642802,0.0025928216,0.042715736,0.00017879649,-0.003360813,-0.030811848,-0.025867647,0.0003514107,0.01620463,0.0075534326,-0.019507354,-0.01807688,0.03619628,-0.022567049,0.0060769864,0.02735956,-0.005774099]",{"tags":37,"relatedLang":46,"relatedPosts":50},[38,39,41,43,44],{"name":17,"slug":17},{"name":13,"slug":40},"kimi-k26",{"name":14,"slug":42},"gpt-55",{"name":16,"slug":16},{"name":15,"slug":45},"swe-bench-pro",{"id":28,"slug":47,"title":48,"language":49},"why-kimi-k26-changes-coding-model-race-en","Why Kimi K2.6 Changes the Coding Model Race","en",[51,57,63,69,75,81],{"id":52,"slug":53,"title":54,"cover_image":55,"image_url":55,"created_at":56,"category":27},"c2abd58c-029c-4e1e-97cc-8f5a5ca969e2","mistral-cybersecurity-model-banks-europe-zh","Mistral 要做銀行資安模型","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778999023148-7ikg.png","2026-05-17T06:23:22.680619+00:00",{"id":58,"slug":59,"title":60,"cover_image":61,"image_url":61,"created_at":62,"category":27},"cf56a4be-6a68-4053-b262-0b343406b995","kimi-k2-6-complete-guide-2026-zh","Kimi K2.6 2026 變了什麼","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778981033614-xldp.png","2026-05-17T01:23:31.568906+00:00",{"id":64,"slug":65,"title":66,"cover_image":67,"image_url":67,"created_at":68,"category":27},"bd8cfc0e-66db-4546-9b9e-fa328f7538d6","weishenme-google-yincang-de-gemini-live-moxing-bi-yanshi-gen-zh","為什麼 Google 隱藏的 Gemini Live 模型，比演示更重要","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778869245574-c25w.png","2026-05-15T18:20:23.111559+00:00",{"id":70,"slug":71,"title":72,"cover_image":73,"image_url":73,"created_at":74,"category":27},"5b5fa24f-5259-4e9e-8270-b08b6805f281","minimax-m1-open-hybrid-attention-reasoning-model-zh","MiniMax-M1：開源 1M Token 推理模型","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778797859209-ea1g.png","2026-05-14T22:30:38.636592+00:00",{"id":76,"slug":77,"title":78,"cover_image":79,"image_url":79,"created_at":80,"category":27},"b1da56ac-8019-4c6b-a8dc-22e6e22b1cb5","gemini-omni-video-review-text-rendering-zh","Gemini Omni 影片模型怎麼了","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778779280109-lrrk.png","2026-05-14T17:20:42.608312+00:00",{"id":82,"slug":83,"title":84,"cover_image":85,"image_url":85,"created_at":86,"category":27},"d63e9d93-e613-4bbf-8135-9599fde11d08","why-xiaomi-mimo-v25-pro-changes-coding-agents-zh","為什麼 Xiaomi 的 MiMo-V2.5-Pro 改變的是 Coding …","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778689858139-v38e.png","2026-05-13T16:30:27.893951+00:00",[88,93,98,103,108,113,118,123,128,133],{"id":89,"slug":90,"title":91,"created_at":92},"58b64033-7eb6-49b9-9aab-01cf8ae1b2f2","nvidia-rubin-six-chips-one-ai-supercomputer-zh","NVIDIA Rubin 把六顆晶片塞進 AI 機櫃","2026-03-26T07:18:45.861277+00:00",{"id":94,"slug":95,"title":96,"created_at":97},"0dcc2c61-c2a6-480d-adb8-dd225fc68914","march-2026-ai-model-news-what-mattered-zh","2026 年 3 月 AI 模型新聞重點","2026-03-26T07:32:08.386348+00:00",{"id":99,"slug":100,"title":101,"created_at":102},"214ab08b-5ce5-4b5c-8b72-47619d8675dd","why-small-models-are-winning-on-device-ai-zh","小模型為何吃下裝置端 AI","2026-03-26T07:36:30.488966+00:00",{"id":104,"slug":105,"title":106,"created_at":107},"785624b2-0355-4b82-adc3-de5e45eecd88","midjourney-v8-faster-images-higher-costs-zh","Midjourney V8 變快了，也變貴了","2026-03-26T07:52:03.562971+00:00",{"id":109,"slug":110,"title":111,"created_at":112},"cda76b92-d209-4134-86c1-a60f5bc7b128","xiaomi-mimo-trio-agents-robots-voice-zh","小米 MiMo 三模型瞄準代理、機器人與語音","2026-03-28T03:05:08.779489+00:00",{"id":114,"slug":115,"title":116,"created_at":117},"9e1044b4-946d-47fe-9e2a-c2ee032e1164","xiaomi-mimo-v2-pro-1t-moe-agents-zh","小米 MiMo-V2-Pro 登場：1T MoE 模型","2026-03-28T03:06:19.002353+00:00",{"id":119,"slug":120,"title":121,"created_at":122},"d68e59a2-55eb-4a8f-95d6-edc8fcbff581","cursor-composer-2-started-from-kimi-zh","Cursor Composer 2 其實從 Kimi 起步","2026-03-28T03:11:58.893796+00:00",{"id":124,"slug":125,"title":126,"created_at":127},"c4b6186f-bd84-4598-997e-c6e31d543c0d","cursor-composer-2-agentic-coding-model-zh","Cursor Composer 2 走向代理式寫碼","2026-03-28T03:13:06.422716+00:00",{"id":129,"slug":130,"title":131,"created_at":132},"45812c46-99fc-4b1f-aae1-56f64f5c9024","openai-shuts-down-sora-video-app-api-zh","OpenAI 關閉 Sora App 與 API","2026-03-29T04:47:48.974108+00:00",{"id":134,"slug":135,"title":136,"created_at":137},"e112e76f-ec3b-408f-810e-e93ae21a888a","apple-siri-gemini-distilled-models-zh","Apple Siri 牽手 Gemini 的真相","2026-03-29T04:52:57.886544+00:00"]