[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-detail":3},{"article":4,"alternate":35,"related":40,"latest":49},{"orderNumber":5,"showImage":6,"title":7,"metaDescription":8,"jumpUrl":9,"content":10,"metaKeywords":11,"withAllowSearch":12,"modified":13,"viewCount":14,"id":15,"lang":16,"slug":17,"dpTemplateId":18,"thumbnail":6,"withHot":19,"withLeadNews":19,"author":20,"created":13,"withTop":19,"highlightContent":21,"userId":22,"highlightTitle":7,"commentStatus":12,"commentCount":5,"thumbnailToContent":12,"withRecommend":19,"metaTitle":23,"editMode":24,"siteId":5,"user":25,"authorEn":20,"status":33,"categoryId":34,"summary":28},0,"\u002Fattachment\u002F20260928\u002F85ac3f9f4cdf4ad1952749eca863c5dd.webp","DeepSeek's Thinking Upgrade: One Model, Two Minds","DeepSeek's online upgrade gives every user thinking and non-thinking modes, weeks after V4.1 Flash retired its own flagship.","https:\u002F\u002Fpoly-ai.chat\u002Fmediasync-claw","\u003Cp style=\"margin:0 0 18px;\">No press conference. No demo video. Late Monday night, DeepSeek simply announced that its online models had been upgraded — and that the new version ships \u003Cstrong>two modes in one model\u003C\u002Fstrong>: a \u003Cstrong>thinking mode\u003C\u002Fstrong> and a \u003Cstrong>non-thinking mode\u003C\u002Fstrong>, available to users right now.\u003C\u002Fp>\n\u003Cp style=\"margin:0 0 18px;\">Most people will read that as a small changelog line. It is not. It is the quiet end of a two-week story in which the smallest model in DeepSeek's new lineup did something unusual: \u003Cstrong>it retired the company's own flagship.\u003C\u002Fstrong>\u003C\u002Fp>\n\u003Ch2 style=\"color:#111;font-size:21px;line-height:1.4;margin:28px 0 12px;\">\u003Cstrong>The Smallest Model That Killed the Flagship\u003C\u002Fstrong>\u003C\u002Fh2>\n\u003Cp style=\"margin:0 0 18px;\">On September 10, DeepSeek released \u003Cstrong>V4.1 Flash\u003C\u002Fstrong>. In the release notes, the company made a claim that rarely appears in a model launch: V4.1 Flash, it said, surpasses \u003Cstrong>V4 Pro on performance, cost, speed, and total time\u003C\u002Fstrong> — and V4 Pro would be retired.\u003C\u002Fp>\n\u003Cp style=\"margin:0 0 18px;\">A Flash model. The \"cheap, fast\" tier. Outperforming the Pro tier across the board, then watching the Pro tier get shut down.\u003C\u002Fp>\n\u003Cp style=\"margin:0 0 18px;\">The benchmark sheet backs the claim. GPQA Diamond: \u003Cstrong>90.9\u003C\u002Fstrong>. Codeforces rating: \u003Cstrong>3471\u003C\u002Fstrong>. MathArena Apex: \u003Cstrong>65.6\u003C\u002Fstrong>. DeepSWE v1.1: \u003Cstrong>74.2\u003C\u002Fstrong>. Terminal-Bench 2.1: \u003Cstrong>90.6\u003C\u002Fstrong>. CyberGym: \u003Cstrong>88.1\u003C\u002Fstrong>. Vision-heavy agent benchmarks like Chartography and BabyVision land at 78.9 and 89.6 with tools — because this is the first Flash with \u003Cstrong>native multimodal visual understanding\u003C\u002Fstrong>. It does not just read text; it looks at images, charts, and interfaces and acts on what it sees.\u003C\u002Fp>\n\u003Cp style=\"margin:0 0 18px;\">All of it on a new asymmetric architecture built for a higher capability ceiling, faster inference, higher throughput — and lower prices. Old Flash models were retired too, with their names temporarily routed to V4.1 Flash so nobody's code breaks.\u003C\u002Fp>\n\u003Cp style=\"margin:0 0 18px;\">\u003Cimg src=\"\u002Fattachment\u002F20260928\u002F653d4999042a49018627628f43ed15d8.webp\" alt=\"DeepSeek V4.1 Flash — smallest model outranks and retires the flagship\">\u003C\u002Fp>\n\u003Ch2 style=\"color:#111;font-size:21px;line-height:1.4;margin:28px 0 12px;\">\u003Cstrong>One Model, Two Minds\u003C\u002Fstrong>\u003C\u002Fh2>\n\u003Cp style=\"margin:0 0 18px;\">Now the upgrade lands at the product level: the same model offers \u003Cstrong>thinking\u003C\u002Fstrong> and \u003Cstrong>non-thinking\u003C\u002Fstrong> behavior, and ordinary users — the people who just open the app and type — can switch between them.\u003C\u002Fp>\n\u003Cp style=\"margin:0 0 18px;\">Thinking mode is the reasoning path: it plans, checks, and works through the problem before answering. Non-thinking mode is the fast path: straight answer, low latency, ideal for the daily bulk of requests that do not need a chain of thought.\u003C\u002Fp>\n\u003Cp style=\"margin:0 0 18px;\">This is not a gimmick split. DeepSeek has been building toward it for months. Since the V4 generation went GA in August, thinking effort has been adjustable in \u003Cstrong>three levels — low, high, and max\u003C\u002Fstrong> — so a task can choose how long it thinks before it speaks. And with V3.2, released in late September, DeepSeek became the first open model to fold thinking into \u003Cstrong>tool calling itself\u003C\u002Fstrong>: the model can reason about which tool to use, and why, while it is using it — earning gold in IMO, CMO, ICPC, and IOI along the way.\u003C\u002Fp>\n\u003Cp style=\"margin:0 0 18px;\">Put together, the direction is clear: thinking used to be a premium toggle for power users. Now it is on by default for everyone, with an off switch for the cases that do not need it.\u003C\u002Fp>\n\u003Cp style=\"margin:0 0 18px;\">\u003Cimg src=\"\u002Fattachment\u002F20260928\u002Fc8e814e82ae446bbb647310890eb5683.webp\" alt=\"DeepSeek thinking and non-thinking modes — winding thought path versus straight fast path\">\u003C\u002Fp>\n\u003Ch2 style=\"color:#111;font-size:21px;line-height:1.4;margin:28px 0 12px;\">\u003Cstrong>What It Actually Changes for You\u003C\u002Fstrong>\u003C\u002Fh2>\n\u003Cp style=\"margin:0 0 18px;\">For users: the app you already have now answers questions it previously fumbled — the ones that need a model to stop, think, and reconsider before replying.\u003C\u002Fp>\n\u003Cp style=\"margin:0 0 18px;\">For developers: prices went down with the V4.1 Flash release, the old model names still resolve, and one API call now controls how much thinking a task gets.\u003C\u002Fp>\n\u003Cp style=\"margin:0 0 18px;\">For the open-source world: this is a reminder of how DeepSeek keeps redefining what \"cheap\" means. Two weeks ago, the company's smallest model outran its flagship and retired it. This week, thinking mode — the feature rivals sell as a premium tier — became a default in a free chat app.\u003C\u002Fp>\n\u003Cp style=\"margin:0 0 18px;\">The changelog line was short. The story under it is two weeks of a company making its flagship obsolete on purpose, then handing the upgrade to every user for free.\u003C\u002Fp>\n\u003Cdiv class=\"dp-template-card\" style=\"border-radius:8px;box-shadow:0 2px 8px rgba(0,0,0,0.1);margin:10px 0;max-width:100%;overflow:hidden;width:100%;\">\n \u003Ca style=\"display:block;text-decoration:none;\" href=\"https:\u002F\u002Fpoly-ai.chat\u002Fmediasync-claw\" target=\"_blank\">\u003Cimg class=\"image_resized\" style=\"display:block;height:auto;max-width:100%;width:100%;\" src=\"\u002Fattachment\u002F20260824\u002Fdb1f58e4d6e24f2f8dcccc811df6e1a8.png\" alt=\"db1f58e4d6e24f2f8dcccc811df6e1a8\">\n  \u003Cbutton style=\"align-items:center;background-color:#784fe2;border-radius:0 0 4px 4px;border-style:none;color:#ffffff;cursor:pointer;display:flex;font-family:Times New Roman;font-size:20px;height:40px;justify-content:center;padding:0;width:100%;\">Experience Now\u003C\u002Fbutton>\u003C\u002Fa>\n\u003C\u002Fdiv>\n\u003Cp style=\"color:#8a8a8a;font-size:13px;margin:0 0 10px;\">\u003Ci>References:\u003C\u002Fi>\u003C\u002Fp>\n\u003Cp style=\"color:#8a8a8a;font-size:13px;margin:0 0 6px;\">\u003Ci>[1] DeepSeek release notes — DeepSeek-V4.1-Flash release and API changes, api-docs.deepseek.com\u002Fupdates, 2026-09-10.\u003C\u002Fi>\u003C\u002Fp>\n\u003Cp style=\"color:#8a8a8a;font-size:13px;margin:0 0 6px;\">\u003Ci>[2] DeepSeek official news — DeepSeek-V4.1 Flash: stronger, faster, more accessible, deepseek.com\u002Fnews, 2026-09-10.\u003C\u002Fi>\u003C\u002Fp>\n\u003Cp style=\"color:#8a8a8a;font-size:13px;margin:0 0 6px;\">\u003Ci>[3] DeepSeek V4 technical documentation — model card and architecture, fe-static.deepseek.com, 2026-04-27.\u003C\u002Fi>\u003C\u002Fp>\n\u003Cp style=\"color:#8a8a8a;font-size:13px;margin:0 0 6px;\">\u003Ci>[4] DeepSeek-V3.2 official release notes — thinking in tool calling, api-docs.deepseek.com, 2025-12-01.\u003C\u002Fi>\u003C\u002Fp>\n\u003Cp style=\"color:#8a8a8a;font-size:13px;margin:0 0 6px;\">\u003Ci>[5] All illustrations are AI-generated.\u003C\u002Fi>\u003C\u002Fp>","yunpoly, aipollo, mediasync-claw, DeepSeek, thinking mode, V4.1 Flash, open source LLM",true,"2026-09-28 23:50:13",92,12599,"zh_CN","deepseek-thinking-upgrade-nxqt",23,false,"yunpoly","No press conference. No demo video. Late Monday night, DeepSeek simply announced that its online mod...",18,"DeepSeek's Thinking Upgrade: One Model, Two Minds | yunpoly","html",{"emailStatusOk":19,"statusLocked":19,"mobileStatusOk":19,"englishNickname":26,"nickname":26,"statusReg":19,"id":22,"created":27,"sourceString":28,"avatar":29,"url":30,"statusOk":19,"detailUrl":31,"username":32},"Science Guide Wwai","2025-12-29 12:22:32","","\u002Fstatic\u002Fcommons\u002Fimg\u002Favatar.png","\u002Fuser\u002F18","\u002Fadmin\u002Fuser\u002Fdetail\u002F18","panhb","normal",72,{"id":36,"slug":37,"title":38,"lang":39},12637,"c69497-lactose-intolerance-gene-benefit-diabetes-risk","乳糖不耐竟是「基因紅利」？研究揭祕：拉肚子竟能防糖尿病","zh",{"prev":41,"next":45},{"id":42,"slug":43,"title":44,"categoryId":34},12601,"jev-judgment-model-zvrn","Jev AI Model: Judgment Without Words",{"id":46,"slug":47,"title":48,"categoryId":34},12598,"mimo-open-source-rl","MiMo V2.6: The Open-Source Model That Beat Grok",[50,55,60,65],{"id":51,"slug":52,"title":53,"thumbnail":54,"categoryId":34},12664,"2f0b3c-ai-snake-romance-gender-debate","When the Fiancée Is a Snake: AI Anime Sparks Debate on Gender Roles and Absurd Romance","https:\u002F\u002Fcdn.banyunjuhe.com\u002Fattachment\u002F20260930\u002Fabd307865c604c1fa501ee8fe6cfff73.png",{"id":56,"slug":57,"title":58,"thumbnail":59,"categoryId":34},12662,"b86dc5-ai-scheming-doubao-debate","AI Auctions Spark Debate: Is 'Doubao' the Ultimate Schemer?","https:\u002F\u002Fcdn.banyunjuhe.com\u002Fattachment\u002F20260930\u002Fcb0922ad05854444856a5e6ff36be9b5.png",{"id":61,"slug":62,"title":63,"thumbnail":64,"categoryId":34},12660,"b5e00a-ai-short-drama-tutorial-controversy","The Rise of the AI \"Evil Cultivator\": A New Era of Digital Content Creation","https:\u002F\u002Fcdn.banyunjuhe.com\u002Fattachment\u002F20260930\u002Fdb61bf4e6fd74f36844203afb950ce59.png",{"id":66,"slug":67,"title":68,"thumbnail":69,"categoryId":34},12658,"0cdfe4-ai-supercar-design-apollo-bugatti-ferrari","Can AI-Designed Supercars Outshine Legacy Brands?","https:\u002F\u002Fcdn.banyunjuhe.com\u002Fattachment\u002F20260930\u002F67262dae9f7048cb8ad5d44f5a0ce3e5.png"]