[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-april-2026-open-source-ai-projects-watch-zh":3,"tags-april-2026-open-source-ai-projects-watch-zh":35,"related-lang-april-2026-open-source-ai-projects-watch-zh":51,"related-posts-april-2026-open-source-ai-projects-watch-zh":55,"series-industry-4e82e9ad-4f0d-449f-b769-aa7035d4ffd4":92},{"id":4,"title":5,"content":6,"summary":7,"source":8,"source_url":9,"author":10,"image_url":11,"keywords":12,"language":23,"translated_content":10,"views":24,"is_premium":25,"created_at":26,"updated_at":26,"cover_image":11,"published_at":27,"rewrite_status":28,"rewrite_error":10,"rewritten_from_id":29,"slug":30,"category":31,"related_article_id":32,"status":33,"google_indexed_at":34,"x_posted_at":10,"tweet_text":10,"title_rewritten_at":10,"title_original":10,"key_takeaways":10,"topic_cluster_id":10,"embedding":10,"is_canonical_seed":25},"4e82e9ad-4f0d-449f-b769-aa7035d4ffd4","2026年4月值得追的開源 AI 專案","\u003Cp>2026 年 4 月的開源 AI 很吵。\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-python\" target=\"_blank\" rel=\"noopener\">Google ADK for Python\u003C\u002Fa> 上線兩週就破 8,200 顆星。\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex-cli\" target=\"_blank\" rel=\"noopener\">OpenAI Codex CLI\u003C\u002Fa> 也衝到 5,800 顆星。\u003C\u002Fp>\u003Cp>模型端更誇張。\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fmeta-llama\u002FLlama-4-Scout-17B\" target=\"_blank\" rel=\"noopener\">Llama-4-Scout-17B\u003C\u002Fa> 在 \u003Ca href=\"https:\u002F\u002Fhuggingface.co\" target=\"_blank\" rel=\"noopener\">Hugging Face\u003C\u002Fa> 一週就拿到 120 萬次下載。講白了，開源 AI 現在不是只有熱鬧，是真的在改變開發者\u003Ca href=\"\u002Fnews\u002Fbeijing-march-15-sandstorm-protection-guide-zh\">怎麼\u003C\u002Fa>做產品。\u003C\u002Fp>\u003Cp>這波最有意思的地方，不是誰聲量最大。是大家開始押同一件事：agent 框架、程式碼模型、本地推論、還有「一上線就有 weights 和可跑程式」的發佈方式。這些東西很務實，也很台灣工程師會在意。\u003C\u002Fp>\u003Ch2>GitHub 上真正吸睛的專案\u003C\u002Fh2>\u003Cp>4 月的 GitHub 很像工具箱大亂鬥。大家不太買單空泛 demo。真正有聲量的，多半是能直接塞進工作流的基礎工具。像是 agent、文件處理、模型部署，這些才是開發者每天會碰到的痛點。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1776211622123-luij.png\" alt=\"2026年4月值得追的開源 AI 專案\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-python\" target=\"blank\" rel=\"noopener\">Google ADK\u003C\u002Fa> 是這波最亮眼的專案之一。兩週 8,200+ stars，不是小數字。它主打多 agent 系統，對想做任務協作、流程自動化的人很實用。\u003C\u002Fp>\u003Cp>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fmeta-llama\u002Fllama-stack\" target=\"_blank\" rel=\"noopener\">Llama Stack\u003C\u002Fa> 也衝到 6,400+ stars。\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex-cli\" target=\"_blank\" rel=\"noopener\">Codex CLI\u003C\u002Fa> 則有 5,800+ stars。這三個名字放一起，很明顯：agent、部署、終端機內寫 code，現在就是主戰場。\u003C\u002Fp>\u003Cul>\u003Cli>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-python\" target=\"_blank\" rel=\"noopener\">Google ADK\u003C\u002Fa>：8,200+ stars，Python，多 agent 系統\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fmeta-llama\u002Fllama-stack\" target=\"_blank\" rel=\"noopener\">Llama Stack\u003C\u002Fa>：6,400+ stars，Python，Llama 4 部署工具\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex-cli\" target=\"_blank\" rel=\"noopener\">Codex CLI\u003C\u002Fa>：5,800+ stars，TypeScript，沙箱化 coding agent\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fblock\u002Fgoose\" target=\"_blank\" rel=\"noopener\">Goose\u003C\u002Fa>：4,900+ stars，Rust，本地優先 agent 框架\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Fsmolagents\" target=\"_blank\" rel=\"noopener\">smolagents\u003C\u002Fa>：4,100+ stars，Python，輕量工具型 agents\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fmarkitdown\" target=\"_blank\" rel=\"noopener\">MarkItDown\u003C\u002Fa>：3,600+ stars，Python，文件轉 Markdown\u003C\u002Fli>\u003C\u002Ful>\u003Cp>我覺得 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fmarkitdown\" target=\"_blank\" rel=\"noopener\">MarkItDown\u003C\u002Fa> 最值得提。它不是那種看起來很炫的模型專案，但它解的是超常見問題。PDF、Office、雜七雜八檔案，最後都要變成乾淨文字。這種工具看起來樸素，卻常常是 LLM 專案的入口。\u003C\u002Fp>\u003Ch2>Hugging Face 的數字代表什麼\u003C\u002Fh2>\u003Cp>\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fmeta-llama\u002FLlama-4-Scout-17B\" target=\"_blank\" rel=\"noopener\">Llama-4-Scout-17B\u003C\u002Fa> 一週 120 萬次下載，這不是單純的流量噱頭。它代表開發者真的想先抓模型回去試。不是看簡報，而是直接跑。\u003C\u002Fp>\u003Cp>\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen3-72B\" target=\"_blank\" rel=\"noopener\">Qwen3-72B\u003C\u002Fa> 也有 64 萬+ 下載。\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fmistralai\u002FCodestral-2-22B\" target=\"_blank\" rel=\"noopener\">Codestral-2-22B\u003C\u002Fa> 則到 38 萬+。這些數字放在一起看，很明顯：程式碼、推理、通用聊天，三條線都還在搶市場。\u003C\u002Fp>\u003Cp>還有一個重點是硬體門檻。\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fmeta-llama\u002FLlama-4-Scout-17B\" target=\"_blank\" rel=\"noopener\">Llama-4-Scout-17B\u003C\u002Fa> 主打 17B active paramet\u003Ca href=\"\u002Fnews\u002Flayer-2-blockchain-scalability-explained-zh\">er\u003C\u002Fa>s，單張 48GB GPU 就能跑。這對台灣很多團隊很重要，因為不是每家公司都有大叢集，也不是每個案子都值得燒雲端費用。\u003C\u002Fp>\u003Cblockquote>“The future is already here — it’s just not evenly distributed.” — William Gibson\u003C\u002Fblockquote>\u003Cp>這句話拿來看 2026 年 4 月，真的很準。最好的開源模型不再躲在論文裡。它們已經可以下載、量化、部署，甚至在一些原本不夠力的硬體上跑起來。\u003C\u002Fp>\u003Cul>\u003Cli>\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fmeta-llama\u002FLlama-4-Scout-17B\" target=\"_blank\" rel=\"noopener\">Llama-4-Scout-17B\u003C\u002Fa>：120 萬+ 下載，單張 48GB GPU 可跑\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen3-72B\" target=\"_blank\" rel=\"noopener\">Qwen3-72B\u003C\u002Fa>：64 萬+ 下載，主打高階推理\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fmistralai\u002FCodestral-2-22B\" target=\"_blank\" rel=\"noopener\">Codestral-2-22B\u003C\u002Fa>：38 萬+ 下載，Apache 2.0 授權\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fgoogle\u002Fgemma-3-9b\" target=\"_blank\" rel=\"noopener\">Gemma-3-9b\u003C\u002Fa>：31 萬+ 下載，商業使用範圍更清楚\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Funsloth\u002FLlama-4-Scout-GGUF\" target=\"_blank\" rel=\"noopener\">Unsloth Llama-4-Scout-GGUF\u003C\u002Fa>：25 萬+ 下載，4-bit 量化格式\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>MoE 為什麼變成主流\u003C\u002Fh2>\u003Cp>4 月最明顯的技術趨勢，就是 Mixture-of-Exp\u003Ca href=\"\u002Fnews\u002Fcertik-opens-ai-auditor-to-global-developers-zh\">ert\u003C\u002Fa>s，簡稱 MoE。以前它像研究圈的特殊玩法。現在它變成很多團隊的標準答案。原因很簡單，大家都想要大模型的效果，但不想每次推論都付密集模型的代價。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1776211634395-eoj6.png\" alt=\"2026年4月值得追的開源 AI 專案\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fdeepseek-ai\u002FDeepSeek-V3-Base\" target=\"_blank\" rel=\"noopener\">DeepSeek V3\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fmeta-llama\u002FLlama-4-Scout-17B\" target=\"_blank\" rel=\"noopener\">Llama 4 Scout\u003C\u002Fa>，還有不少 \u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002FQwen\" target=\"_blank\" rel=\"noopener\">Qwen\u003C\u002Fa> 系列，都用了 MoE 變體。這代表什麼？代表你可能不用養一整排伺服器，才能讓模型看起來夠強。\u003C\u002Fp>\u003Cp>這件事很現實。active parameters 變少，記憶體壓力就小。延遲也會比較好看。對內部 copilot、客服機器人、文件助理這種案子，能不能上線，常常就卡在這裡。\u003C\u002Fp>\u003Cul>\u003Cli>\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fdeepseek-ai\u002FDeepSeek-V3-Base\" target=\"_blank\" rel=\"noopener\">DeepSeek V3 Base\u003C\u002Fa>：671B 總參數，37B active\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen3-Coder-32B\" target=\"_blank\" rel=\"noopener\">Qwen3-Coder-32B\u003C\u002Fa>：128K context，原生 tool calling\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Funslothai\u002Funsloth\" target=\"_blank\" rel=\"noopener\">Unsloth\u003C\u002Fa>：2x faster fine-tuning，70% less memory\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002FHuggingFaceTB\u002FSmolVLM2-2.2B\" target=\"_blank\" rel=\"noopener\">SmolVLM2-2.2B\u003C\u002Fa>：18 萬+ 下載，小型多模態模型\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fblack-forest-labs\u002FFLUX.1-Kontext\" target=\"_blank\" rel=\"noopener\">FLUX.1-Kontext\u003C\u002Fa>：16 萬+ 下載，圖片編輯與文字渲染\u003C\u002Fli>\u003C\u002Ful>\u003Cp>如果你要我直白講，MoE 就是在算力很貴的年代，幫大家把帳單壓低一點。不是免費，但比硬扛 dense model 友善多了。\u003C\u002Fp>\u003Ch2>開發者該先試哪個\u003C\u002Fh2>\u003Cp>如果你是工程師，別先追最紅的。先看你的瓶頸在哪。卡在寫 code，就先碰 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex-cli\" target=\"_blank\" rel=\"noopener\">Codex CLI\u003C\u002Fa> 或 \u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen3-Coder-32B\" target=\"_blank\" rel=\"noopener\">Qwen3-Coder-32B\u003C\u002Fa>。這兩個都很適合做程式輔助。\u003C\u002Fp>\u003Cp>如果你在做 agent 流程，就看 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-python\" target=\"_blank\" rel=\"noopener\">Google ADK\u003C\u002Fa> 和 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fblock\u002Fgoose\" target=\"_blank\" rel=\"noopener\">Goose\u003C\u002Fa>。前者偏多 agent 編排，後者偏本地優先。兩者路線不同，但都比很多只會畫圖的 demo 實用。\u003C\u002Fp>\u003Cp>如果你想做本地推論，量化版本最值得碰。像 \u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Funsloth\u002FLlama-4-Scout-GGUF\" target=\"_blank\" rel=\"noopener\">Unsloth 的 GGUF 版本\u003C\u002Fa>，就是那種會讓部署變簡單的東西。少一堆安裝地獄，團隊才會真的用。\u003C\u002Fp>\u003Cp>看 repo 時，我會先看 issue 區。星星很多，不代表真的有人在用。issue 很活躍，通常才表示它進了實戰。\u003C\u002Fp>\u003Ch2>這波開源 AI 背後的產業脈絡\u003C\u002Fh2>\u003Cp>開源 AI 這幾年有個很明顯的變化。以前大家先丟論文，再慢慢補 code。現在是反過來，先把模型、權重、demo、安裝方式一起丟出來。開發者沒耐心等，產品團隊也沒空陪你猜。\u003C\u002Fp>\u003Cp>這也解釋了為什麼 GitHub 和 Hugging Face 的影響力一直在變大。前者看的是可用工具，後者看的是模型擴散速度。兩邊一起熱，代表一件事：AI 不只在研究室裡跑，也在真實專案裡跑。\u003C\u002Fp>\u003Cp>對台灣團隊來說，這很實際。很多公司沒有無限預算。你要的是能上伺服器、能控成本、能接 API、能跟既有軟體整合的方案。這也是為什麼輕量 agent、本地模型、量化版本，現在特別有吸引力。\u003C\u002Fp>\u003Cp>我也覺得，接下來會有更多團隊把重點放在部署細節，而不是只比參數量。誰能把模型塞進 1 台伺服器，誰能把回應時間壓到可接受，誰就更容易進到產品。\u003C\u002Fp>\u003Ch2>4 月這波，接下來怎麼看\u003C\u002Fh2>\u003Cp>如果要我下個判斷，接下來幾個月會更偏向「能跑」而不是「看起來很強」。能直接接進工作流的專案，會比單純展示能力的專案更吃香。這很現實，也很合理。\u003C\u002Fp>\u003Cp>你如果現在就要挑一個方向，我會先看 agent 框架和量化模型。前者決定工作流怎麼接，後者決定成本能不能壓住。這兩塊做對了，專案才有機會從試玩變成日常工具。\u003C\u002Fp>\u003Cp>所以問題其實很簡單：你現在手上的 AI 專案，是在追星，還是在解問題？如果是後者，4 月這批開源專案，已經給了不少可以直接上手的答案。\u003C\u002Fp>","2026 年 4 月的開源 AI 很熱鬧。GitHub 的 agent 工具、Hugging Face 的模型下載數都很猛，這篇整理最值得看的專案、數據和實際影響。","fazm.ai","https:\u002F\u002Ffazm.ai\u002Fblog\u002Fnew-open-source-ai-projects-github-hugging-face-april-2026",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1776211622123-luij.png",[13,14,15,16,17,18,19,20,21,22],"開源AI","GitHub","Hugging Face","agent 框架","MoE","Llama 4","Codex CLI","Qwen","本地推論","量化模型","zh",2,false,"2026-04-15T00:06:44.711318+00:00","2026-04-15T00:06:44.606+00:00","done","26602516-3a36-4792-aeb6-8dda79f3a017","april-2026-open-source-ai-projects-watch-zh","industry","d69bbb37-b7de-4a9f-ad7f-33874aa1c355","published","2026-04-15T09:00:09.049+00:00",[36,38,40,41,43,45,47,49],{"name":18,"slug":37},"llama-4",{"name":13,"slug":39},"開源ai",{"name":22,"slug":22},{"name":15,"slug":42},"hugging-face",{"name":20,"slug":44},"qwen",{"name":14,"slug":46},"github",{"name":16,"slug":48},"agent-框架",{"name":19,"slug":50},"codex-cli",{"id":32,"slug":52,"title":53,"language":54},"april-2026-open-source-ai-projects-watch-en","April 2026’s Open Source AI Projects Worth Watching","en",[56,62,68,74,80,86],{"id":57,"slug":58,"title":59,"cover_image":60,"image_url":60,"created_at":61,"category":31},"e6379f8a-3305-4862-bd15-1192d3247841","why-nebius-ai-pivot-is-more-real-than-hype-zh","為什麼 Nebius 的 AI 轉型比炒作更真實","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778823044520-9mfz.png","2026-05-15T05:30:24.978992+00:00",{"id":63,"slug":64,"title":65,"cover_image":66,"image_url":66,"created_at":67,"category":31},"66c4e357-d84d-43ef-a2e7-120c4609e98e","nvidia-backs-corning-factories-with-billions-zh","Nvidia 出資 Corning 工廠擴產","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778822450270-trdb.png","2026-05-15T05:20:27.701475+00:00",{"id":69,"slug":70,"title":71,"cover_image":72,"image_url":72,"created_at":73,"category":31},"31d8109c-8b0b-46e2-86bc-d274a03269d1","why-anthropic-gates-foundation-ai-public-goods-zh","為什麼 Anthropic 和 Gates Foundation 應該投資 A…","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778796636474-u508.png","2026-05-14T22:10:21.138177+00:00",{"id":75,"slug":76,"title":77,"cover_image":78,"image_url":78,"created_at":79,"category":31},"17cafb6e-9f2c-43c4-9ba3-ef211d2780b1","why-observability-is-critical-cloud-native-systems-zh","為什麼可觀測性是雲原生系統的生存條件","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778794245143-tfqn.png","2026-05-14T21:30:25.97324+00:00",{"id":81,"slug":82,"title":83,"cover_image":84,"image_url":84,"created_at":85,"category":31},"2fb441af-d3c6-4af8-a356-a40b25a67c00","data-centers-pushing-homeowners-to-solar-zh","資料中心推升房主裝太陽能","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778793651300-gi06.png","2026-05-14T21:20:40.899115+00:00",{"id":87,"slug":88,"title":89,"cover_image":90,"image_url":90,"created_at":91,"category":31},"387bddd8-e5fc-4aa9-8d1b-43a34b0ece43","how-to-choose-gpu-for-yihuan-zh","怎麼選《异环》GPU","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778786461303-39mx.png","2026-05-14T19:20:29.220124+00:00",[93,98,103,108,113,118,123,128,133,138],{"id":94,"slug":95,"title":96,"created_at":97},"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":99,"slug":100,"title":101,"created_at":102},"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":104,"slug":105,"title":106,"created_at":107},"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":109,"slug":110,"title":111,"created_at":112},"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":114,"slug":115,"title":116,"created_at":117},"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":119,"slug":120,"title":121,"created_at":122},"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":124,"slug":125,"title":126,"created_at":127},"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":129,"slug":130,"title":131,"created_at":132},"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":134,"slug":135,"title":136,"created_at":137},"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":139,"slug":140,"title":141,"created_at":142},"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"]