[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-mimo-v2-flash-openrouter-benchmarks-pricing-zh":3,"article-related-mimo-v2-flash-openrouter-benchmarks-pricing-zh":34,"series-model-release-09fe28b5-aae5-4bac-b3bd-9a261e4c99a1":89},{"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":25,"views":30,"created_at":31,"published_at":32,"topic_cluster_id":33},"09fe28b5-aae5-4bac-b3bd-9a261e4c99a1","mimo-v2-flash-openrouter-benchmarks-pricing-zh","MiMo-V2-Flash 直衝開源 SWE-bench","\u003Cp data-speakable=\"summary\">Xiaomi 的 \u003Ca href=\"\u002Fnews\u002Fmimo-v25-pro-turns-agent-work-into-one-api-call-zh\">MiMo\u003C\u002Fa>-V2-Flash 是一款 309B 參數的開源 MoE \u003Ca href=\"\u002Fnews\u002Fgovernment-access-orders-frontier-model-access-zh\">模型\u003C\u002Fa>，OpenRouter 也已列出它的價格與測試表現。\u003C\u002Fp>\u003Cp>這個組合很直接。模型很大，價格卻壓得很低。OpenRouter 顯示它的輸入每 1M Token 是 $0.10，輸出是 $0.30。對開發者來說，這種定價很有殺傷力。\u003C\u002Fp>\u003Cp>更麻煩的是，它還把開源 \u003Ca href=\"\u002Ftag\u002Fswe-bench\">SWE-bench\u003C\u002Fa> 的成績拉到前面。對現在一堆 LLM 來說，會聊天不稀奇。能真的修 code，才是比較像樣的事。\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>項目\u003C\u002Fth>\u003Cth>數值\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>總參數\u003C\u002Ftd>\u003Ctd>309B\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>OpenRouter 輸入價格\u003C\u002Ftd>\u003Ctd>$0.10 \u002F 1M Token\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>OpenRouter 輸出價格\u003C\u002Ftd>\u003Ctd>$0.30 \u002F 1M Token\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>模型類型\u003C\u002Ftd>\u003Ctd>Mixture-of-Experts\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>先講結論，這顆模型很會算帳\u003C\u002Fh2>\u003Cp>先看最現實的部分。\u003Ca href=\"https:\u002F\u002Fopenrouter.ai\u002Fxiaomi\u002Fmimo-v2-flash\" target=\"_blank\" rel=\"noopener\">OpenRouter\u003C\u002Fa> 把 \u003Ca href=\"https:\u002F\u002Fwww.mi.com\u002Fglobal\u002F\" target=\"_blank\" rel=\"noopener\">Xiaomi\u003C\u002Fa> 的 \u003Ca href=\"https:\u002F\u002Fopenrouter.ai\u002Fxiaomi\u002Fmimo-v2-flash\" target=\"_blank\" rel=\"noopener\">MiMo-V2-Flash\u003C\u002Fa> 放上去時，價格直接打到很低。這不是學術海報上的漂亮數字，而是你真的會看到的 API 單價。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781321565467-96el.png\" alt=\"MiMo-V2-Flash 直衝開源 SWE-bench\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>309B 參數聽起來很嚇人。可是 MoE 架構的重點，就是不是每次推論都把全部參數打開。講白了，這種設計就是想在成本和能力之間找平衡。對雲端服務商來說，這很重要。\u003C\u002Fp>\u003Cp>如果你在做軟體產品，成本就是硬傷。尤其是 code assistant、客服自動化、文件摘要這類場景，Token 量很快就爆。每 1M Token 只要 $0.10 \u002F $0.30，代表它很適合拿來跑大量請求。\u003C\u002Fp>\u003Cul>\u003Cli>大模型，但價格壓得低。\u003C\u002Fli>\u003Cli>MoE 架構，推論成本比較好控。\u003C\u002Fli>\u003Cli>適合大量 Token 的產品。\u003C\u002Fli>\u003Cli>對原型驗證很友善。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>SWE-bench 這件事，才是重點\u003C\u002Fh2>\u003Cp>很多模型都會寫詩、會聊天、會講廢話。可是真正難的是修 bug。\u003Ca href=\"https:\u002F\u002Fwww.swebench.com\u002F\" target=\"_blank\" rel=\"noopener\">SWE-bench\u003C\u002Fa> 就是在測這件事。它看模型能不能處理真實 \u003Ca href=\"\u002Ftag\u002Fgithub\">GitHub\u003C\u002Fa> issue，然後改出可用的\u003Ca href=\"\u002Fnews\u002F6-kuan-ai-cheng-shi-dai-li-de-2026-fen-gong-zh\">程式\u003C\u002Fa>碼。\u003C\u002Fp>\u003Cp>MiMo-V2-Flash 在開源 SWE-bench 上衝到前段，這代表它不是只會背答案。它得讀懂 repo 結構、找出錯誤點、再產出能過測試的 patch。這比單純問答難很多。\u003C\u002Fp>\u003Cp>我覺得這種分數很有參考價值。因為開發者在乎的不是模型講得多像人，而是它能不能少浪費你 2 小時。你如果有用過 code \u003Ca href=\"\u002Ftag\u002Fagent\">agent\u003C\u002Fa>，就知道一個錯誤 patch 能把整個流程搞爛。\u003C\u002Fp>\u003Cblockquote>\"The best way to predict the future is to invent it.\" — Alan Kay\u003C\u002Fblockquote>\u003Cp>這句話放在這裡很貼。工具不是拿來看熱鬧的。它要是真的能修 code，才會進到工作流。SWE-bench 的分數，就是這條線上的一個門檻。\u003C\u002Fp>\u003Ch2>和其他模型比，差在哪裡\u003C\u002Fh2>\u003Cp>先說現況。開源 coding 模型很多。\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FQwenLM\u002FQwen3\" target=\"_blank\" rel=\"noopener\">Qwen\u003C\u002Fa> 系列、\u003Ca href=\"https:\u002F\u002Fwww.deepseek.com\u002F\" target=\"_blank\" rel=\"noopener\">DeepSeek\u003C\u002Fa> 系列、\u003Ca href=\"https:\u002F\u002Fai.meta.com\u002Fllama\u002F\" target=\"_blank\" rel=\"noopener\">Llama\u003C\u002Fa> 系列，都在搶這塊市場。大家都想證明自己不只是大，而是真的好用。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781321563721-wsgh.png\" alt=\"MiMo-V2-Flash 直衝開源 SWE-bench\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>MiMo-V2-Flash 的賣點很明確。第一，它是 309B。第二，它走 MoE。第三，它在開源 SWE-bench 上有不錯表現。第四，它的價格很低。這四個條件放一起，就不是單一亮點，而是一整套商業打法。\u003C\u002Fp>\u003Cp>但也別太快高潮。大模型的 \u003Ca href=\"\u002Ftag\u002Fbenchmark\">benchmark\u003C\u002Fa> 很會騙人。今天在 SWE-bench 亮眼，不代表你丟到自己公司的私有 repo，也會一樣順。資料格式、依賴版本、測試習慣，這些都會讓結果走鐘。\u003C\u002Fp>\u003Cul>\u003Cli>\u003Ca href=\"https:\u002F\u002Fopenrouter.ai\u002Fxiaomi\u002Fmimo-v2-flash\" target=\"_blank\" rel=\"noopener\">MiMo-V2-Flash\u003C\u002Fa>：主打低價與 coding 表現。\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fwww.deepseek.com\u002F\" target=\"_blank\" rel=\"noopener\">DeepSeek\u003C\u002Fa>：常被拿來比 code 能力。\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fai.meta.com\u002Fllama\u002F\" target=\"_blank\" rel=\"noopener\">Llama\u003C\u002Fa>：生態成熟，部署選項多。\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FQwenLM\u002FQwen3\" target=\"_blank\" rel=\"noopener\">Qwen\u003C\u002Fa>：中文與 agent 場景存在感高。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>這種價格，會怎麼改變開發流程\u003C\u002Fh2>\u003Cp>如果價格真的長期維持在這個區間，很多團隊會開始改流程。以前可能只把大模型留給高價值任務。現在你可能會想把它塞進更多日常步驟，像是 \u003Ca href=\"\u002Ftag\u002Fcode-review\">code review\u003C\u002Fa>、測試生成、issue 分類。\u003C\u002Fp>\u003Cp>這會讓 AI API 的使用方式更像基礎設施，而不是展示品。講白了，便宜才有機會變成預設選項。貴模型很強，但如果每次都要精算 Token，團隊最後還是會縮手。\u003C\u002Fp>\u003Cp>不過成本低也有代價。你要看延遲、穩定性、上下文長度，還有供應商的服務品質。便宜 API 很香，但如果常常 timeout，工程團隊還是會罵人。這點在生產環境特別明顯。\u003C\u002Fp>\u003Ch2>MiMo-V2-Flash 背後的訊號\u003C\u002Fh2>\u003Cp>這顆模型還有一個訊號很清楚。中國大型硬體與軟體公司，正在把 AI 模型做成完整產品線。不是只有手機、伺服器、雲端服務，現在連開源 LLM 也要一起上。\u003C\u002Fp>\u003Cp>這對台灣開發者不是壞事。市場上多一個便宜又能打 code 的選項，代表你在選模型時有更多籌碼。你可以拿它跟商用閉源模型比，也可以拿它來做內部測試。\u003C\u002Fp>\u003Cp>真正該看的，是它會不會進入更多工具鏈。像 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex\" target=\"_blank\" rel=\"noopener\">OpenAI Codex\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-code\" target=\"_blank\" rel=\"noopener\">Claude Code\u003C\u002Fa> 這類工作流，已經把模型當成工程工具。MiMo-V2-Flash 如果能在價格和表現上同時站穩，就會更容易被接進去。\u003C\u002Fp>\u003Ch2>接下來該盯什麼\u003C\u002Fh2>\u003Cp>我會先看三件事。第一，實際 API 延遲。第二，長上下文穩定度。第三，真實 repo 上的修復成功率。這三項比單一 benchmark 更接近現場。\u003C\u002Fp>\u003Cp>如果你是工程團隊，現在就可以做一件事。拿你們最常見的 bug 類型，做一組小型測試集。把 MiMo-V2-Flash 跟你現在用的模型放一起比。別只看分數，也看人工修正時間。\u003C\u002Fp>\u003Cp>說真的，這種模型最可怕的地方，不是參數多，而是價格低到讓你很難忽視。接下來幾個月，重點不是它會不會被討論，而是有多少團隊真的把它接進 production。","Xiaomi 的 MiMo-V2-Flash 以 309B MoE 架構登場，OpenRouter 標價每 1M Token 只要 $0.10 \u002F $0.30，並在開源 SWE-bench 分數上衝到前段班。","openrouter.ai","https:\u002F\u002Fopenrouter.ai\u002Fxiaomi\u002Fmimo-v2-flash",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781321565467-96el.png","model-release","zh","e7f37851-7b5f-429c-9a71-3e4a2d4b9c70",[17,18,19,20,21,22,23,24],"MiMo-V2-Flash","Xiaomi","OpenRouter","SWE-bench","開源 LLM","MoE","API 價格","code model",[26,27,28,29],"MiMo-V2-Flash 是 309B 參數的開源 MoE 模型，OpenRouter 已列出 API 價格。","它在開源 SWE-bench 的表現，讓它不只是聊天模型，而是 code 工具候選人。","每 1M Token $0.10 \u002F $0.30 的定價，對大量推論場景很有吸引力。","真正要驗證的，還是延遲、穩定性和真實 repo 的修復成功率。",0,"2026-06-13T03:32:17.367685+00:00","2026-06-13T03:32:17.359+00:00","0ccb5d2e-69f1-4354-a3e0-cb370221cd95",{"tags":35,"relatedLang":48,"relatedPosts":52},[36,38,40,42,45],{"name":19,"slug":37},"openrouter",{"name":18,"slug":39},"xiaomi",{"name":21,"slug":41},"開源-llm",{"name":43,"slug":44},"MiMo V2 Flash","mimo-v2-flash",{"name":46,"slug":47},"SWE-Bench","swe-bench",{"id":15,"slug":49,"title":50,"language":51},"mimo-v2-flash-openrouter-benchmarks-pricing-en","MiMo-V2-Flash hits top open-source SWE-bench scores","en",[53,59,65,71,77,83],{"id":54,"slug":55,"title":56,"cover_image":57,"image_url":57,"created_at":58,"category":13},"f26334ab-dd8b-49c2-a49e-7fc376200f2b","microsoft-bets-on-controllable-domain-tuned-models-zh","微軟押注可控、領域調校模型，而不是更大的通用模型","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781331468190-ymfp.png","2026-06-13T06:17:20.311904+00:00",{"id":60,"slug":61,"title":62,"cover_image":63,"image_url":63,"created_at":64,"category":13},"0bb91791-f4b6-4d51-899c-6eeb239f942a","microsoft-mai-models-build-2026-zh","Microsoft把 Copilot 拉回主場","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781330585064-s9ya.png","2026-06-13T06:02:35.160901+00:00",{"id":66,"slug":67,"title":68,"cover_image":69,"image_url":69,"created_at":70,"category":13},"eaafd0fe-cb56-40cd-80f8-c203e3d72f03","gpt-5-4-thinkng-pro-mini-nano-release-zh","GPT-5.4 率先登場，mini、nano 跟進","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781326973205-xufk.png","2026-06-13T05:02:19.576989+00:00",{"id":72,"slug":73,"title":74,"cover_image":75,"image_url":75,"created_at":76,"category":13},"0a9dbc64-2e51-494d-b6b6-21ecfd8dd1f5","minimax-m3-1m-token-coding-power-zh","MiniMax M3 把 1M Token 送進寫碼場","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781295477857-78hl.png","2026-06-12T20:17:28.037784+00:00",{"id":78,"slug":79,"title":80,"cover_image":81,"image_url":81,"created_at":82,"category":13},"8ca34749-2efa-4b24-b2bd-c0fb66062b49","openai-confidential-ipo-us-stock-market-zh","OpenAI 低調送件 IPO","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781294579049-ltqt.png","2026-06-12T20:02:33.419196+00:00",{"id":84,"slug":85,"title":86,"cover_image":87,"image_url":87,"created_at":88,"category":13},"b15b0887-bd5b-43e6-ac42-23939d0f4e92","google-gemini-35-pro-june-2m-token-launch-zh","Gemini 3.5 Pro 6月登場，2M Token 夠猛","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781204585839-bdsh.png","2026-06-11T19:02:36.371587+00:00",[90,95,100,105,110,115,120,125,130,135],{"id":91,"slug":92,"title":93,"created_at":94},"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":96,"slug":97,"title":98,"created_at":99},"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":101,"slug":102,"title":103,"created_at":104},"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":106,"slug":107,"title":108,"created_at":109},"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":111,"slug":112,"title":113,"created_at":114},"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":116,"slug":117,"title":118,"created_at":119},"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":121,"slug":122,"title":123,"created_at":124},"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":126,"slug":127,"title":128,"created_at":129},"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",{"id":131,"slug":132,"title":133,"created_at":134},"c679b51f-194a-463b-87fc-7695256ff752","mimo-v2-pro-vs-omni-vs-flash-2026-zh","MiMo V2 Pro、Omni、Flash 怎麼選","2026-04-02T01:18:43.576128+00:00",{"id":136,"slug":137,"title":138,"created_at":139},"3b988fd7-6749-4f01-ba25-c0ad7486dc31","z-ai-glm-5v-turbo-design2code-claude-zh","GLM-5V-Turbo 在 Design2Code 贏了…","2026-04-02T04:03:36.31741+00:00"]