[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"content-doc-dcioecd27c1d":3},{"user":4,"document":8,"mainDocument":27,"columnUrl":29,"subscription":30,"footer":42,"text":80},{"isAuthenticated":5,"isAdmin":5,"displayName":6,"avatarUrl":6,"nid":6,"groupLevel":7},false,"",-10,{"id":9,"fullTitle":10,"subTitle":6,"url":11,"columnId":12,"columnName":13,"columnUrl":14,"summary":6,"contentHtml":15,"mainContentHtml":6,"posterUrl":16,"createDate":17,"displayDate":18,"displayDateSlash":19,"pageviews":20,"tags":21,"hidden":5,"isSubContent":5,"replyDocOrTargetId":6,"contentType":23,"videoId":6,"liveVideoUrl":6,"useContentVideo":5,"duration":24,"price":24,"priceText":25,"priceBadgeText":25,"priceBadgeClass":26,"freeForMinGroupLevel":24,"redirectUrl":6,"readyToStream":5},"dcioecd27c1d","Kimi 创始人：详细披露了Kimi技术路线","\u002Fdoc\u002Fdcioecd27c1d","col18178739ee","美股资讯","\u002Fcol\u002Fcol18178739ee","\u003Cp>\u003Cspan style=\"font-size: large;\">中国 Kimi 大模型背后的 CEO 表示：AI 竞赛的关键并不是谁拥有最多资源。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">“真正稀缺的资源不是芯片，而是洞察力（Insights）。”\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">Moonshot AI（月之暗面）创始人杨植麟详细披露了他们如何依靠开源模型与 GPT、Claude 正面竞争的技术路线。在美国头部实验室逐渐减少技术公开分享的时候，他们选择了另一条道路。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">第一步，他们放弃了整个行业自 2014 年以来广泛使用的 Adam 优化器。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">Moonshot AI 开发的 Muon 优化器变体，可以用约一半的 FLOPs（浮点运算量）达到相同的训练损失（Loss），同时实现两倍的 Token 效率。这意味着在相同算力条件下，模型能够学习更多内容，或者用更少资源达到同样效果。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">第二步，他们优化的不仅仅是模型损失函数（Loss），而是“每个 Token 位置对应的 Loss”。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">杨植麟认为，优秀的模型架构不应该随着上下文长度增加而逐渐失去能力，而应该在更长的上下文中持续提升理解和推理能力。而这恰恰是 AI Agent 时代最重要的竞争核心——谁能在超长上下文中保持更强的推理和执行能力，谁就更有优势。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">第三步，他们解决了大规模训练过程中长期存在的不稳定性问题。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">在超大规模训练时，Loss 曲线经常会出现突然飙升（Spike），导致训练效率下降甚至失败。Moonshot AI 采用了一项名为 QK-Clip 的技术，在长达 15 万亿 Token 的训练过程中，让 Loss 曲线始终保持平稳，没有出现任何异常波动。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">杨植麟甚至表示，这是他在 2025 年见过“最美丽”的技术成果之一。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">第四步，他们正在推进下一代模型 K3 所使用的 Kimi Linear 技术。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">这是一种线性注意力（Linear Attention）机制。长期以来，线性注意力虽然理论上效率更高，但效果始终无法超越传统的全注意力（Full Attention）架构。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">而 Moonshot AI 认为，他们终于实现了突破。在处理超长文本任务时，Kimi Linear 的性能已经超过传统全注意力机制，并且在百万 Token 上下文场景下，速度最高可提升 6 倍。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">第五步，开源并不是一种慷慨行为，而是一种战略。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">杨植麟认为，每一次开源发布都会推动整个行业建立在他们的技术成果之上。随着越来越多开发者和企业采用这些模型，中国开源大模型正在逐渐成为全球 AI 生态中的默认选择之一。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">据美股投资网 TradesMax.com ，最后，他用一句话概括了这一切：\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">“构建一个模型，本质上是在构建一种世界观（Worldview）。”\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">而到了最后，真正决定模型高度的，往往不是参数规模，也不是算力投入，而是创造者的品味（Taste）。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">观看并收藏这段分享。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp>\u003Cspan style=\"font-size: large;\">这是迄今为止最清晰的一次机会，让人们了解 Kimi 团队究竟是如何思考 AI，以及他们为何能够在资源远不如美国科技巨头的情况下，持续与 GPT、Claude 等顶级模型展开竞争。\u003C\u002Fspan>\u003C\u002Fp>\r\n\u003Cp> \u003C\u002Fp>","https:\u002F\u002Fwww.tradesmax.com\u002Fimages\u002Fa_Stock\u002FA\u002FAI\u002FAI.jpg","2026-07-22T22:08:36","2026.07.22","2026\u002F07\u002F22",41512,[22],"AI","Article",0,"免费","success",{"id":9,"fullTitle":10,"subTitle":6,"url":11,"columnId":12,"columnName":13,"columnUrl":14,"summary":6,"contentHtml":15,"mainContentHtml":6,"posterUrl":16,"createDate":17,"displayDate":18,"displayDateSlash":19,"pageviews":20,"tags":28,"hidden":5,"isSubContent":5,"replyDocOrTargetId":6,"contentType":23,"videoId":6,"liveVideoUrl":6,"useContentVideo":5,"duration":24,"price":24,"priceText":25,"priceBadgeText":25,"priceBadgeClass":26,"freeForMinGroupLevel":24,"redirectUrl":6,"readyToStream":5},[22],"\u002Fcol\u002Fstocknews",{"visible":5,"marketingHtml":31,"services":32,"recentDocuments":41},"\u003Cfigure class=\"image\">\u003Ca href=\"https:\u002F\u002Fstockwe.com\u002Fdoc\u002Fdcio537efad5\" target=\"_blank\" rel=\"noopener noreferrer\">\u003Cimg style=\"display:block;margin-left:auto;margin-right:auto;\" src=\"\u002Fimg\u002Fstockwewebfiles\u002Fweb-202408-stk\u002F1586109431mceclip0.jpg\">\u003C\u002Fa>\u003C\u002Ffigure>\u003Cdiv class=\"text-center\">\u003Ch2 class=\"card-title mx-auto\">\u003Cbr>\u003Ca target=\"_blank\" rel=\"noopener noreferrer\" href=\"https:\u002F\u002Fstockwe.com\u002Fdoc\u002Fdcio537efad5\">案例介绍：英伟达深度研究报告\u003C\u002Fa>\u003C\u002Fh2>\u003C\u002Fdiv>",[33,37],{"productId":34,"serviceName":35,"priceText":36},"prod_PPxdDdK87QaiLv","月付","$12.95美元",{"productId":38,"serviceName":39,"priceText":40},"prod_PPxeMs3bix1da5","年付","$149.00美元",[],{"links":43,"images":71,"summaryHtml":76,"aboutTitle":77,"aboutHtml":78,"copyrightHtml":79},[44,47,50,53,56,59,62,65,68],{"label":45,"url":46},"深度报告","\u002Fcol\u002FdepthReport",{"label":48,"url":49},"VIP会员","\u002Fvip",{"label":51,"url":52},"期权推荐","\u002FOption",{"label":54,"url":55},"低价暴涨股","\u002FPenny",{"label":57,"url":58},"AI智能体","\u002FAiAgent",{"label":60,"url":61},"常见问题","https:\u002F\u002Fstockwe.com\u002FFAQ",{"label":63,"url":64},"美股课程","\u002Fcol\u002Fvideos",{"label":66,"url":67},"免责声明","\u002Fdisclaimer",{"label":69,"url":70},"联系我们","\u002FContactUs",[72,73,74,75],"\u002Fimg\u002Fstockwebsiteblob\u002Fweb-202509-stk\u002FUploaderzic2tuwsol2_2025_09_11_18_21_07.gif","\u002Fimg\u002Fstockwebsiteblob\u002Fweb-202509-stk\u002FUploadercakzdvydksw_2025_09_03_09_00_56.png","\u002Fimg\u002Fstockwebsiteblob\u002Fweb-202509-stk\u002FUploadergtjyagwvoyk_2025_09_14_08_32_05.png","\u002Fimg\u002Fstockwebsiteblob\u002Fweb-202509-stk\u002FUploader3u0tt4jhlqh_2025_09_23_22_30_48.png","邮箱: buy@TradesMax.com 美国电话 626-378-3637","公司介绍","\u003Cp class=\"MsoNormal\">美股大数据 \u003Ca href=\"https:\u002F\u002Fstockwe.com\" target=\"_blank\" rel=\"noopener\">StockWe.com\u003C\u002Fa> 是一个美国领先的金融和美股信息大数据提供商，紧盯华尔街金融市场和行情，2008年成立于美国硅谷，创始人是前纽约证券交易所资深分析师Ken，联合多位摩根斯坦利分析师，谷歌 Meta工程师利用AI和大数据，配合十多年美股实战经验和业内量化交易模型，每天处理海量股票数据：挖掘潜力大牛股，捕捉期权异动大单，实时主力资金流向、机构持仓变化、川普突发新闻，美股买卖信号第一时间发到您手机APP。\u003C\u002Fp>","专业美股投资者都在这里",{"loading":81,"search":82,"searchPlaceholder":82,"hotContent":83,"draft":84,"noData":85,"searchNoData":86,"edit":87,"editVideo":88,"courseContent":89,"more":90,"buyNow":91,"subscribeNow":92,"encoding":93,"paidContent":94},"Loading...","搜索","热门内容","草稿","目前没有任何内容公布","当前检索内容没有数据","编辑","编辑视频","课程内容","更多","立即购买后观看","- 立即订阅 -","视频编码中...","付费内容"]