[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-tether-bitnet-fine-tuning-edge-devices-zh":3,"article-related-tether-bitnet-fine-tuning-edge-devices-zh":30,"series-model-release-8e754dee-26eb-443d-8766-1cc31a4522bd":82},{"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":22,"views":26,"created_at":27,"published_at":28,"topic_cluster_id":29},"8e754dee-26eb-443d-8766-1cc31a4522bd","tether-bitnet-fine-tuning-edge-devices-zh","Tether 推 Bitnet 邊緣微調","\u003Cp data-speakable=\"summary\">Tether 於 2026 年 5 月 29 日發布 Bitnet \u003Ca href=\"\u002Fnews\u002Fcode2lora-repo-specific-adapters-code-models-en-zh\">LoRA\u003C\u002Fa> 框架，主打可在消費級裝置上微調 13B 模型。\u003C\u002Fp>\u003Cp>Tether 這次把焦點放在邊緣裝置，不再只依賴雲端 \u003Ca href=\"\u002Ftag\u002Fgpu\">GPU\u003C\u002Fa>。官方說法是，手機、筆電、桌機都能參與訓練與推理，讓 AI 開發更靠近使用者手上的硬體。\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>項目\u003C\u002Fth>\u003Cth>數值\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>發布日期\u003C\u002Ftd>\u003Ctd>2026 年 5 月 29 日\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>模型規模\u003C\u002Ftd>\u003Ctd>130 億參數\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>每週生成式 AI 使用者\u003C\u002Ftd>\u003Ctd>約 7 億\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>大型公司 AI 擴張率\u003C\u002Ftd>\u003Ctd>近 50%\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>小型公司 AI 擴張率\u003C\u002Ftd>\u003Ctd>29%\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>發生了什麼\u003C\u002Fh2>\u003Cp>\u003Ca href=\"https:\u002F\u002Ftether.io\" target=\"_blank\" rel=\"noopener\">Tether\u003C\u002Fa> 公布的框架，\u003Ca href=\"\u002Fnews\u002Flinux-kernel-hobby-project-core-infrastructure-zh\">核心\u003C\u002Fa>是把 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FBitNet\" target=\"_blank\" rel=\"noopener\">Microsoft Bitnet\u003C\u002Fa> 延伸到 LoRA 微調。它支援異質消費級 GPU，包含行動 GPU，也把執行後端擴到 Vulkan 與 Metal。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780729377117-xoyl.png\" alt=\"Tether 推 Bitnet 邊緣微調\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>這代表 Bitnet 不再只侷限於原本的 Bitnet.cpp 推理路徑。Tether 也加入 dynamic tiling，專門處理行動硬體常見的 driver buffer 限制，避免裝置端微調卡在記憶體配置上。\u003C\u002Fp>\u003Cp>官方同時把這套能力包進 \u003Ca href=\"https:\u002F\u002Fqvac.dev\" target=\"_blank\" rel=\"noopener\">QVAC SDK\u003C\u002Fa>。Tether 表示，開發者可以把它用在 QVAC Workbench，並把工作分配到不同裝置上，形成比較完整的 edge-first 工具鏈。\u003C\u002Fp>\u003Cul>\u003Cli>支援 Vulkan 與 Metal GPU\u003C\u002Fli>\u003Cli>可在手機、PC、筆電上跑\u003C\u002Fli>\u003Cli>把推理與 LoRA 微調放到同一套框架\u003C\u002Fli>\u003Cli>以開源方式提供給開發者\u003C\u002Fli>\u003C\u002Ful>\u003Cp>文章提到的目標裝置很具體，包括 Samsung S25 與 iPhone 16 等級的手機，以及一般個人電腦。這意味著 Tether 想證明，130 億參數模型不一定只能在\u003Ca href=\"\u002Ftag\u002F資料中心\">資料中心\u003C\u002Fa>裡調整。\u003C\u002Fp>\u003Cp>這套做法也不是憑空出現。Tether 說 dynamic tiling 先前已用在 QVAC Fabric \u003Ca href=\"\u002Ftag\u002Fllm\">LLM\u003C\u002Fa> fine-tuning framework，現在只是把同一思路搬到 Bitnet 上，讓更多硬體型態能接上來。\u003C\u002Fp>\u003Ch2>為什麼重要\u003C\u002Fh2>\u003Cp>對開發者來說，最大變化是成本結構。若微調能在本機完成，小團隊就不必為每次實驗都租大型 GPU 叢集，原型開發、客製化與測試都會更快。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780729366861-kc26.png\" alt=\"Tether 推 Bitnet 邊緣微調\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>這對零售、SMB 與消費型應用特別有用。這些場景常常\u003Ca href=\"\u002Fnews\u002Fwhy-ai-needs-a-brake-pedal-now-zh\">需要\u003C\u002Fa>依照地端資料或個人習慣調整模型，但又不想把敏感資料送上雲端，裝置端微調就成了更直接的選項。\u003C\u002Fp>\u003Cp>產業面上，這也碰上 AI 擴張速度的落差。文章引用 McKinsey 2025 State of AI 調查，指出營收超過 50 億美元的公司，近 50% 已進入 AI 擴張階段；營收低於 1 億美元的公司，比例只有 29%。Tether 的賭注是，把算力放回使用者裝置，能縮小這道差距。\u003C\u002Fp>\u003Cp>它還把 \u003Ca href=\"https:\u002F\u002Fholepunch.to\" target=\"_blank\" rel=\"noopener\">Holepunch\u003C\u002Fa>、Pear 與 delegated \u003Ca href=\"\u002Ftag\u002Finference\">inference\u003C\u002Fa> 串進同一個敘事裡。重點不是單一模型，而是讓裝置之間直接協作，形成不依賴雲端的分散式應用模式。\u003C\u002Fp>\u003Cp>真正的考題也很直白：消費級 GPU、行動驅動與開源工具，能不能撐住可重複、可上線的微調流程？如果答案是可以，邊緣 AI 的門檻會明顯下降；如果不行，這仍會先停留在示範層。\u003C\u002Fp>\u003Cp>對產業來說，這篇消息的重點不是 Bitnet 本身，而是 Tether 正在把「模型訓練」從雲端專案改寫成裝置功能。\u003C\u002Fp>","Tether 於 2026 年 5 月 29 日發布 Bitnet LoRA 微調框架，主打在手機、筆電與桌機上跑 13B 模型，並把訓練與推理往邊緣裝置移動。","www.computerworld.com","https:\u002F\u002Fwww.computerworld.com\u002Farticle\u002F4177577\u002Fdemocratizing-ai-adoption-with-tethers-bitnet-llm-fine-tuning-framework.html",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780729377117-xoyl.png","model-release","zh","d9b6ff74-204d-41d8-a118-669ead54dba0",[17,18,19,20,21],"Tether","Bitnet","LoRA","邊緣 AI","裝置端微調",[23,24,25],"Tether 把 Bitnet LoRA 微調帶到消費級裝置，目標是手機、筆電與桌機。","框架支援 Vulkan、Metal 與 dynamic tiling，重點在降低行動硬體的限制。","若本機微調可落地，小團隊與中小型應用會少掉一大筆 GPU 成本。",0,"2026-06-06T07:02:26.208165+00:00","2026-06-06T07:02:26.2+00:00","0ccb5d2e-69f1-4354-a3e0-cb370221cd95",{"tags":31,"relatedLang":41,"relatedPosts":45},[32,34,36,38,40],{"name":19,"slug":33},"lora",{"name":20,"slug":35},"邊緣-ai",{"name":17,"slug":37},"tether",{"name":18,"slug":39},"bitnet",{"name":21,"slug":21},{"id":15,"slug":42,"title":43,"language":44},"tether-bitnet-fine-tuning-edge-devices-en","Tether's Bitnet fine-tuning brings AI to edge devices","en",[46,52,58,64,70,76],{"id":47,"slug":48,"title":49,"cover_image":50,"image_url":50,"created_at":51,"category":13},"29e59d4e-6ccc-422b-afdb-18290e6fe168","best-open-source-llms-2026-zh","2026 最強開源 LLM 清單","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780731186899-0vr7.png","2026-06-06T07:32:37.635885+00:00",{"id":53,"slug":54,"title":55,"cover_image":56,"image_url":56,"created_at":57,"category":13},"0392d382-6364-45bc-8532-8e6759930499","mips-risc-v-ai-ip-ces-edge-models-zh","MIPS 推出 RISC-V 邊緣 AI IP","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780668189122-x05o.png","2026-06-05T14:02:32.582526+00:00",{"id":59,"slug":60,"title":61,"cover_image":62,"image_url":62,"created_at":63,"category":13},"003f59ba-8d5a-40cb-8e6b-0b51898bc537","midjourney-21-second-video-model-closed-ai-wrong-deal-zh","為什麼 Midjourney 的 21 秒影片模型證明封閉式 AI 是錯的交易","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780648396657-oa18.png","2026-06-05T08:32:37.400103+00:00",{"id":65,"slug":66,"title":67,"cover_image":68,"image_url":68,"created_at":69,"category":13},"b5889da8-fa42-44ed-89a7-3347655b388d","microsoft-seven-ai-models-openai-anthropic-build-2026-zh","7 款 Microsoft AI 模型登場","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780642975404-f8mr.png","2026-06-05T07:02:23.607092+00:00",{"id":71,"slug":72,"title":73,"cover_image":74,"image_url":74,"created_at":75,"category":13},"b5926931-ce20-4b9d-8814-a3c960187209","what-we-know-about-gpt-56-release-date-zh","GPT-5.6 何時發布？目前線索整理","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780574585815-dzo7.png","2026-06-04T12:02:35.122398+00:00",{"id":77,"slug":78,"title":79,"cover_image":80,"image_url":80,"created_at":81,"category":13},"1985ce38-03c6-4968-96fa-b751553bbef3","why-claude-opus-48-is-not-the-big-story-zh","為什麼 Claude Opus 4.8 不是大新聞","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780531367297-nrfs.png","2026-06-04T00:02:24.633987+00:00",[83,88,93,98,103,108,113,118,123,128],{"id":84,"slug":85,"title":86,"created_at":87},"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":89,"slug":90,"title":91,"created_at":92},"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":94,"slug":95,"title":96,"created_at":97},"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":99,"slug":100,"title":101,"created_at":102},"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":104,"slug":105,"title":106,"created_at":107},"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":109,"slug":110,"title":111,"created_at":112},"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":114,"slug":115,"title":116,"created_at":117},"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":119,"slug":120,"title":121,"created_at":122},"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":124,"slug":125,"title":126,"created_at":127},"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":129,"slug":130,"title":131,"created_at":132},"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"]