[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-google-gemini-35-pro-june-2m-token-launch-zh":3,"article-related-google-gemini-35-pro-june-2m-token-launch-zh":33,"series-model-release-b15b0887-bd5b-43e6-ac42-23939d0f4e92":85},{"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":29,"created_at":30,"published_at":31,"topic_cluster_id":32},"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","\u003Cp data-speakable=\"summary\">\u003Ca href=\"\u002Ftag\u002Fgoogle\">Google\u003C\u002Fa> 計畫在 6 月推出 \u003Ca href=\"\u002Ftag\u002Fgemini\">Gemini\u003C\u002Fa> 3.5 Pro，重點是 2M Token 上下文長度，適合長文件、程式碼庫和多輪分析。\u003C\u002Fp>\u003Cp>說真的，2M Token 不是小數字。這種上下文長度，已經不是一般聊天機器人的玩法了。\u003C\u002Fp>\u003Cp>你可能會想問，這到底有多大。簡單講，這會直接影響資料分析、程式除錯，還有長篇文件整理。\u003C\u002Fp>\u003Cp>如果 \u003Ca href=\"https:\u002F\u002Fblog.google\u002Ftechnology\u002Fai\u002Fgoogle-io-2026\u002F\" target=\"_blank\" rel=\"noopener\">Google I\u002FO\u003C\u002Fa> 真的把這件事端上來，開發者圈子一定會很有感。\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>項目\u003C\u002Fth>\u003Cth>數字\u003C\u002Fth>\u003Cth>意義\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>Gemini 3.5 Pro\u003C\u002Ftd>\u003Ctd>2M Token\u003C\u002Ftd>\u003Ctd>可吃進更長的上下文\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>預計時間\u003C\u002Ftd>\u003Ctd>6 月\u003C\u002Ftd>\u003Ctd>產品節奏很快\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>Gemini 1.5 Pro\u003C\u002Ftd>\u003Ctd>1M Token\u003C\u002Ftd>\u003Ctd>先前版本的常見對照\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>OpenAI GPT-4.1\u003C\u002Ftd>\u003Ctd>1M Token\u003C\u002Ftd>\u003Ctd>同級競品常拿來比\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>2M Token 到底有什麼用\u003C\u002Fh2>\u003Cp>先講白話。上下文長度，就是\u003Ca href=\"\u002Fnews\u002Fmlx-community-apple-silicon-model-weights-zh\">模型\u003C\u002Fa>一次能記住多少內容。Token 不是字數，但你可以把它想成模型能同時看的資料量。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781204585839-bdsh.png\" alt=\"Gemini 3.5 Pro 6月登場，2M Token 夠猛\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>2M Token 的價值，不在炫技。它真正有用的地方，是長文件、長對話、長程式碼庫。\u003C\u002Fp>\u003Cp>像是法務合約、研究報告、產品規格書，這些東西常常很長。以前你要分段餵給模型，現在可以少切很多次。\u003C\u002Fp>\u003Cul>\u003Cli>可以直接丟更長的 PDF 或文件集\u003C\u002Fli>\u003Cli>可處理大型程式碼庫的跨檔案關聯\u003C\u002Fli>\u003Cli>適合做多輪分析，不用一直重複背景\u003C\u002Fli>\u003Cli>對摘要、檢索、比對任務很實用\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>Google 這次在拚什麼\u003C\u002Fh2>\u003Cp>Google 很清楚，LLM 不是只有會聊天就好。真正能留住開發者的，是速度、成本、上下文長度，還有 API 穩定性。\u003C\u002Fp>\u003Cp>\u003Ca href=\"https:\u002F\u002Fdeepmind.google\u002Ftechnologies\u002Fgemini\u002F\" target=\"_blank\" rel=\"noopener\">Gemini\u003C\u002Fa> 系列一直在長上下文這條路上硬拚。這點很像 Google 的老習慣，直接把伺服器和模型容量往上拉。\u003C\u002Fp>\u003Cp>但我覺得重點不只是在 2M Token。更關鍵的是，這個能力會不會真的進到 \u003Ca href=\"https:\u002F\u002Fai.google.dev\u002F\" target=\"_blank\" rel=\"noopener\">Google AI Studio\u003C\u002Fa> 和 API，而且價格別太誇張。\u003C\u002Fp>\u003Cblockquote>“We are at the beginning of a new era in computing.” — Sundar Pichai\u003C\u002Fblockquote>\u003Cp>這句話是 \u003Ca href=\"https:\u002F\u002Fabc.xyz\u002Finvestor\u002F\" target=\"_blank\" rel=\"noopener\">Sundar Pichai\u003C\u002Fa> 說的。雖然很像科技公司標準台詞，但放在這裡還算貼切。\u003C\u002Fp>\u003Cp>因為現在的競爭，不是誰模型名字比較帥。是誰能讓開發者真的拿去做事，而且不會半路爆掉。\u003C\u002Fp>\u003Ch2>和競品比，差在哪裡\u003C\u002Fh2>\u003Cp>如果只看數字，2M Token 很吸睛。但實務上，大家還是會拿它跟 \u003Ca href=\"\u002Ftag\u002Fclaude\">Claude\u003C\u002Fa>、GPT 系列一起比。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781204584658-0c51.png\" alt=\"Gemini 3.5 Pro 6月登場，2M Token 夠猛\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>\u003Ca href=\"https:\u002F\u002Fwww.anthropic.com\u002Fclaude\" target=\"_blank\" rel=\"noopener\">Claude\u003C\u002Fa> 在文件整理和長文推理很有口碑。\u003Ca href=\"https:\u002F\u002Fopenai.com\u002Fgpt-4-1\" target=\"_blank\" rel=\"noopener\">GPT-4.1\u003C\u002Fa> 則在工具整合和生態上很強。Google 的優勢，是它有搜尋、雲端和 Android 這些入口。\u003C\u002Fp>\u003Cp>所以這場不是單純比上下文長度。是比整個產品線能不能把模型能力接到真實工作流。\u003C\u002Fp>\u003Cul>\u003Cli>Gemini：主打超長上下文與 Google 生態\u003C\u002Fli>\u003Cli>Claude：長文理解和寫作體驗很穩\u003C\u002Fli>\u003Cli>GPT 系列：工具鏈和開發者社群很大\u003C\u002Fli>\u003Cli>真正勝負：價格、延遲、API 穩定度\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>開發者會先感受到什麼\u003C\u002Fh2>\u003Cp>第一個感受，\u003Ca href=\"\u002Fnews\u002Fweb3-wallets-financial-super-apps-zh\">應該\u003C\u002Fa>是文件處理。你可以更少切 chunk，少做一些很煩的前處理。這對 RAG 和內部知識庫很重要。\u003C\u002Fp>\u003Cp>第二個感受，是除錯和程式碼\u003Ca href=\"\u002Fnews\u002Feu-defi-review-vagueness-into-policy-zh\">審查\u003C\u002Fa>。當模型能同時看到更多檔案，跨模組追問題會順很多。\u003C\u002Fp>\u003Cp>第三個感受，可能是成本壓力。上下文變長，通常也代表 Token 成本和延遲壓力會更高。這不是魔法，伺服器還是要算錢。\u003C\u002Fp>\u003Cul>\u003Cli>RAG 切塊數可減少\u003C\u002Fli>\u003Cli>長文件摘要更完整\u003C\u002Fli>\u003Cli>跨檔案程式碼分析更方便\u003C\u002Fli>\u003Cli>API 成本可能成為門檻\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>這波背後的產業脈絡\u003C\u002Fh2>\u003Cp>長上下文競爭，已經是 LLM 的基本戰場。以前大家比的是參數量，現在更常比能吃多少資料、能不能穩定輸出。\u003C\u002Fp>\u003Cp>這對\u003Ca href=\"\u002Ftag\u002F台灣開發者\">台灣開發者\u003C\u002Fa>也有感。很多團隊在做客服、知識管理、法遵、電商內容，資料量都不小。模型如果能一次吃更多內容，工作流程會簡化很多。\u003C\u002Fp>\u003Cp>但也別太快高潮。上下文長，不代表答案就一定準。模型還是可能漏看重點，或在長文件裡抓錯方向。\u003C\u002Fp>\u003Cp>所以實際導入時，還是要看評測。像是 \u003Ca href=\"https:\u002F\u002Fai.google.dev\u002Fgemini-api\u002Fdocs\" target=\"_blank\" rel=\"noopener\">Gemini API 文件\u003C\u002Fa>、內部 \u003Ca href=\"\u002Ftag\u002Fbenchmark\">benchmark\u003C\u002Fa>、還有真實工作資料，這些都比宣傳稿重要。\u003C\u002Fp>\u003Ch2>接下來我會看什麼\u003C\u002Fh2>\u003Cp>如果 Gemini 3.5 Pro 真的在 6 月上線，我會先看三件事。第一是價格。第二是延遲。第三是它在真實長文件任務上的準確率。\u003C\u002Fp>\u003Cp>2M Token 很吸睛，但不是唯一答案。能不能真的幫開發者省時間，才是重點。\u003C\u002Fp>\u003Cp>我自己的判斷很直接：如果 Google 把這個能力做進 API，還把價格壓得合理，那它會很有競爭力。反過來，如果只剩數字漂亮，實際用起來慢又貴，那就只是新聞稿好看而已。\u003C\u002Fp>\u003Cp>接下來半年，建議你先準備兩種測試資料。第一種是長文件。第二種是多檔案程式碼庫。等 API 開放後，直接拿真實情境去跑，答案會比空談更清楚。\u003C\u002Fp>","Google 傳出要在 6 月推出 Gemini 3.5 Pro，主打 2M Token 上下文。這代表長文件、程式碼庫和多輪分析會更好處理，但實際表現還是要看價格、速度和穩定性。","www.techtimes.com","https:\u002F\u002Fwww.techtimes.com\u002Farticles\u002F317919\u002F20260606\u002Fgoogle-gemini-35-pro-nears-june-launch-2-million-token-context-deep-think-reasoning.htm",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781204585839-bdsh.png","model-release","zh","3a291fde-d57b-49c2-b0c5-f795b0853c2b",[17,18,19,20,21,22,23,24],"Gemini 3.5 Pro","Google","2M Token","LLM","長上下文","人工智慧","API","程式碼分析",[26,27,28],"Gemini 3.5 Pro 傳出 6 月推出，主打 2M Token 上下文。","2M Token 對長文件、程式碼庫和多輪分析很有用。","真正的重點還是價格、延遲、穩定性和實際準確率。",2,"2026-06-11T19:02:36.371587+00:00","2026-06-11T19:02:36.359+00:00","0a3b4f35-7be1-430e-b708-37bdc8b5219a",{"tags":34,"relatedLang":44,"relatedPosts":48},[35,37,39,41,42],{"name":19,"slug":36},"2m-token",{"name":17,"slug":38},"gemini-35-pro",{"name":20,"slug":40},"llm",{"name":21,"slug":21},{"name":18,"slug":43},"google",{"id":15,"slug":45,"title":46,"language":47},"google-gemini-35-pro-june-2m-token-launch-en","Google Gemini 3.5 Pro Targets June With 2M Tokens","en",[49,55,61,67,73,79],{"id":50,"slug":51,"title":52,"cover_image":53,"image_url":53,"created_at":54,"category":13},"42ca8c4e-e593-461b-b108-ec98c12cf678","unsloth-kimi-k25-gguf-hugging-face-zh","Unsloth 把 Kimi-K2.5 做成 GGUF 包","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781160488625-q93d.png","2026-06-11T06:47:33.607859+00:00",{"id":56,"slug":57,"title":58,"cover_image":59,"image_url":59,"created_at":60,"category":13},"8c573682-2528-4882-bff0-e1a06cd8f2ee","gpt-56-chasing-front-end-before-beating-mythos-zh","GPT-5.6先追前端，再談超越 Mythos","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781154168441-ovuw.png","2026-06-11T05:02:21.52852+00:00",{"id":62,"slug":63,"title":64,"cover_image":65,"image_url":65,"created_at":66,"category":13},"a9be565a-5861-4371-898d-20b98794be42","claude-mythos-5-5000-zh","Claude Mythos 5：一天搬完5000萬行程式","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781148791055-zocy.png","2026-06-11T03:32:40.554558+00:00",{"id":68,"slug":69,"title":70,"cover_image":71,"image_url":71,"created_at":72,"category":13},"4fde468d-be9e-4013-a2e0-8b68ab4bf250","claude-fable-5-quiet-ai-release-week-zh","Claude Fable 5 讓這週像在降溫","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781143383988-o40t.png","2026-06-11T02:02:38.955757+00:00",{"id":74,"slug":75,"title":76,"cover_image":77,"image_url":77,"created_at":78,"category":13},"ef44efd1-dfaf-4d9e-8772-3a6d6f963f08","mistral-model-lineup-specialization-beats-giant-model-zh","Mistral 的模型陣容證明：專精勝過一個巨型模型","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781140675776-0e88.png","2026-06-11T01:17:28.295033+00:00",{"id":80,"slug":81,"title":82,"cover_image":83,"image_url":83,"created_at":84,"category":13},"19af5701-87e3-4774-be7a-8aebcbeef2a5","xiaomi-mimo-1t-model-1000-tokens-per-second-zh","小米 MiMo 把 1T 模型推到 1000 tokens\u002Fs","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781129889723-wz61.png","2026-06-10T22:17:35.161841+00:00",[86,91,96,101,106,111,116,121,126,131],{"id":87,"slug":88,"title":89,"created_at":90},"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":92,"slug":93,"title":94,"created_at":95},"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":97,"slug":98,"title":99,"created_at":100},"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":102,"slug":103,"title":104,"created_at":105},"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":107,"slug":108,"title":109,"created_at":110},"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":112,"slug":113,"title":114,"created_at":115},"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":117,"slug":118,"title":119,"created_at":120},"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":122,"slug":123,"title":124,"created_at":125},"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":127,"slug":128,"title":129,"created_at":130},"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":132,"slug":133,"title":134,"created_at":135},"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"]