社区
首页
集团介绍
社区
资讯
行情
学堂
TigerAI
登录
注册
charmby
炒股养家的码农
IP属地:北京
+关注
帖子 · 782
帖子 · 782
关注 · 0
关注 · 0
粉丝 · 0
粉丝 · 0
charmby
charmby
·
07-17
Mark长个教训 还没熟悉就开始大仓位玩期权 [流泪] [流泪] 玩股票一季度好不容易攒了4倍。赚的钱二季度开始玩期权还回去了 。Mark下 先去攒点本金再回来 放了个大烟花 一定要做回去💪
看
1,457
回复
2
点赞
2
编组 21备份 2
分享
举报
charmby
charmby
·
09-12
mark
阿里DeepSeek时刻!开源新架构模型:推理快10倍、成本暴降90%
阿里巴巴开源Qwen3-Next-80B-A3B新架构模型,融合门控DeltaNet和门控注意力的混合架构,训练成本较Qwen3-32B暴降90%,推理效率提升10倍,在超长文本32K以上场景表现尤佳。性能上,指令微调版本媲美旗舰Qwen3-235B,思考模型超越谷歌Gemini-2.5-Flash,成为最强低能耗开源模型之一。
阿里DeepSeek时刻!开源新架构模型:推理快10倍、成本暴降90%
看
185
回复
评论
点赞
点赞
编组 21备份 2
分享
举报
charmby
charmby
·
08-19
$Strategy(MSTR)$
我靠 这货期权没救了 跌一个点期权就跌50个点 涨一个点 期权涨幅为0。。这样的谁在期权上赚过钱啊
看
2,050
回复
1
点赞
2
编组 21备份 2
分享
举报
charmby
charmby
·
08-18
$Strategy(MSTR)$
这个怎么回事。。。拉起来一半到0.5 居然call创新低????
看
602
回复
评论
点赞
点赞
编组 21备份 2
分享
举报
charmby
charmby
·
07-11
我开仓了55手
$NVDA 20250718 160.0 PUT$
,来看看我最新分享的订单!
看
668
回复
评论
点赞
点赞
编组 21备份 2
分享
举报
charmby
charmby
·
07-01
我平仓了63手
$HOOD 20250711 87.0 PUT$
,要多倒霉有多倒霉[流泪] [流泪] [流泪] aoao
看
759
回复
评论
点赞
点赞
编组 21备份 2
分享
举报
charmby
charmby
·
07-01
我平仓了55手
$NVDA 20250711 157.5 PUT$
,每次都被另一只坑 导致每次节奏都乱 tnd
看
909
回复
评论
点赞
点赞
编组 21备份 2
分享
举报
charmby
charmby
·
06-30
原本备受看好的投资标的(美元、美股等)表现惨淡, 来搞笑了 别人历史新高
半年过去了,华尔街的“脸都被打肿了”
特朗普关税政策和地缘政治冲突彻底颠覆了华尔街年初预测。美元遭遇2005年来最差年初表现,标普500经历惊人暴跌和闪电式反弹。欧洲股市从投资洼地变为必备资产,基准指数跑赢标普500达16个百分点,新兴市场打破连年跑输魔咒,货币普遍兑美元走强,股市为股东创造1.8万亿美元财富增长。
半年过去了,华尔街的“脸都被打肿了”
看
858
回复
评论
点赞
点赞
编组 21备份 2
分享
举报
charmby
charmby
·
06-28
我平仓了47手
$NVDA 20250711 157.5 PUT$
,来看看我最新分享的订单!
看
1,064
回复
评论
点赞
点赞
编组 21备份 2
分享
举报
charmby
charmby
·
06-20
我平仓了108手
$NVDA 20250627 140.0 PUT$
,!有病 开盘诱多 忍不住全割了 fcccccck
看
721
回复
评论
点赞
1
编组 21备份 2
分享
举报
加载更多
暂无关注
热议股票
{"i18n":{"language":"zh_CN"},"isCurrentUser":false,"userPageInfo":{"id":"3494390725694021","uuid":"3494390725694021","gmtCreate":1531297188172,"gmtModify":1749221178826,"name":"charmby","pinyin":"charmby","introduction":"","introductionEn":null,"signature":"炒股养家的码农","avatar":"https://static.tigerbbs.com/c0555fde187648aa2f83f9fe88db109b","hat":null,"hatId":null,"hatName":null,"vip":1,"status":2,"fanSize":244,"headSize":62,"tweetSize":782,"questionSize":0,"limitLevel":900,"accountStatus":4,"level":{"id":0,"name":"","nameTw":"","represent":"","factor":"","iconColor":"","bgColor":""},"themeCounts":1,"badgeCounts":0,"badges":[],"moderator":false,"superModerator":false,"manageSymbols":null,"badgeLevel":null,"boolIsFan":false,"boolIsHead":false,"favoriteSize":0,"symbols":null,"coverImage":null,"realNameVerified":null,"userBadges":[{"badgeId":"35ec162348d5460f88c959321e554969-3","templateUuid":"35ec162348d5460f88c959321e554969","name":"传说交易员","description":"证券或期货账户累计交易次数达到300次","bigImgUrl":"https://static.tigerbbs.com/656db16598a0b8f21429e10d6c1cb033","smallImgUrl":"https://static.tigerbbs.com/03f10910d4dd9234f9b5702a3342193a","grayImgUrl":"https://static.tigerbbs.com/0c767e35268feb729d50d3fa9a386c5a","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2024.12.06","exceedPercentage":"93.94%","individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1100},{"badgeId":"228c86a078844d74991fff2b7ab2428d-3","templateUuid":"228c86a078844d74991fff2b7ab2428d","name":"投资合伙人虎","description":"证券账户累计交易金额达到100万美元","bigImgUrl":"https://static.tigerbbs.com/fbeac6bb240db7da8b972e5183d050ba","smallImgUrl":"https://static.tigerbbs.com/436cdf80292b99f0a992e78750ac4e3a","grayImgUrl":"https://static.tigerbbs.com/506a259a7b456f037592c3b23c779599","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2024.11.09","exceedPercentage":"93.04%","individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1101},{"badgeId":"976c19eed35f4cd78f17501c2e99ef37-1","templateUuid":"976c19eed35f4cd78f17501c2e99ef37","name":"博闻投资者","description":"累计交易超过10只正股","bigImgUrl":"https://static.tigerbbs.com/e74cc24115c4fbae6154ec1b1041bf47","smallImgUrl":"https://static.tigerbbs.com/d48265cbfd97c57f9048db29f22227b0","grayImgUrl":"https://static.tigerbbs.com/76c6d6898b073c77e1c537ebe9ac1c57","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2024.11.01","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1102},{"badgeId":"518b5610c3e8410da5cfad115e4b0f5a-1","templateUuid":"518b5610c3e8410da5cfad115e4b0f5a","name":"实盘交易者","description":"完成一笔实盘交易","bigImgUrl":"https://static.tigerbbs.com/2e08a1cc2087a1de93402c2c290fa65b","smallImgUrl":"https://static.tigerbbs.com/4504a6397ce1137932d56e5f4ce27166","grayImgUrl":"https://static.tigerbbs.com/4b22c79415b4cd6e3d8ebc4a0fa32604","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":1,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2024.10.08","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1100},{"badgeId":"e50ce593bb40487ebfb542ca54f6a561-4","templateUuid":"e50ce593bb40487ebfb542ca54f6a561","name":"明星虎友","description":"加入老虎社区2000天","bigImgUrl":"https://static.tigerbbs.com/dddf24b906c7011de2617d4fb3f76987","smallImgUrl":"https://static.tigerbbs.com/53d58ad32c97254c6f74db8b97e6ec49","grayImgUrl":"https://static.tigerbbs.com/6304700d92ad91c7a33e2e92ec32ecc1","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2024.10.07","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1001}],"userBadgeCount":5,"currentWearingBadge":{"badgeId":"518b5610c3e8410da5cfad115e4b0f5a-1","templateUuid":"518b5610c3e8410da5cfad115e4b0f5a","name":"实盘交易者","description":"完成一笔实盘交易","bigImgUrl":"https://static.tigerbbs.com/2e08a1cc2087a1de93402c2c290fa65b","smallImgUrl":"https://static.tigerbbs.com/4504a6397ce1137932d56e5f4ce27166","grayImgUrl":"https://static.tigerbbs.com/4b22c79415b4cd6e3d8ebc4a0fa32604","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":1,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2024.10.08","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1100},"individualDisplayBadges":null,"crmLevel":9,"crmLevelSwitch":1,"location":"北京","starInvestorFollowerNum":229,"starInvestorFlag":true,"starInvestorOrderShareNum":624,"subscribeStarInvestorNum":27,"ror":-93.57,"winRationPercentage":71.917808,"showRor":true,"investmentPhilosophy":null,"starInvestorSubscribeFlag":false},"page":1,"watchlist":null,"tweetList":[{"id":457622333382656,"gmtCreate":1752761148230,"gmtModify":1752761246351,"author":{"id":"3494390725694021","authorId":"3494390725694021","name":"charmby","avatar":"https://static.tigerbbs.com/c0555fde187648aa2f83f9fe88db109b","crmLevel":9,"crmLevelSwitch":1,"followedFlag":false,"authorIdStr":"3494390725694021","idStr":"3494390725694021"},"themes":[],"htmlText":"Mark长个教训 还没熟悉就开始大仓位玩期权 [流泪] [流泪] 玩股票一季度好不容易攒了4倍。赚的钱二季度开始玩期权还回去了 。Mark下 先去攒点本金再回来 放了个大烟花 一定要做回去💪","listText":"Mark长个教训 还没熟悉就开始大仓位玩期权 [流泪] [流泪] 玩股票一季度好不容易攒了4倍。赚的钱二季度开始玩期权还回去了 。Mark下 先去攒点本金再回来 放了个大烟花 一定要做回去💪","text":"Mark长个教训 还没熟悉就开始大仓位玩期权 [流泪] [流泪] 玩股票一季度好不容易攒了4倍。赚的钱二季度开始玩期权还回去了 。Mark下 先去攒点本金再回来 放了个大烟花 一定要做回去💪","images":[{"img":"https://static.tigerbbs.com/265679e3654573d9fc2e831813cc3e63","width":"1179","height":"2556"}],"top":2,"highlighted":1,"essential":1,"paper":1,"likeSize":2,"commentSize":2,"repostSize":0,"link":"https://laohu8.com/post/457622333382656","isVote":1,"tweetType":1,"viewCount":1457,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":1,"langContent":"CN","totalScore":0},{"id":477717274935768,"gmtCreate":1757640006176,"gmtModify":1757640007835,"author":{"id":"3494390725694021","authorId":"3494390725694021","name":"charmby","avatar":"https://static.tigerbbs.com/c0555fde187648aa2f83f9fe88db109b","crmLevel":9,"crmLevelSwitch":1,"followedFlag":false,"authorIdStr":"3494390725694021","idStr":"3494390725694021"},"themes":[],"htmlText":"mark","listText":"mark","text":"mark","images":[],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":0,"commentSize":0,"repostSize":0,"link":"https://laohu8.com/post/477717274935768","repostId":"2566948680","repostType":4,"repost":{"id":"2566948680","kind":"news","pubTimestamp":1757637309,"share":"https://www.laohu8.com/m/news/2566948680?lang=&edition=full","pubTime":"2025-09-12 08:35","market":"us","language":"zh","title":"阿里DeepSeek时刻!开源新架构模型:推理快10倍、成本暴降90%","url":"https://stock-news.laohu8.com/highlight/detail?id=2566948680","media":"AIGC开放社区","summary":"阿里巴巴开源Qwen3-Next-80B-A3B新架构模型,融合门控DeltaNet和门控注意力的混合架构,训练成本较Qwen3-32B暴降90%,推理效率提升10倍,在超长文本32K以上场景表现尤佳。性能上,指令微调版本媲美旗舰Qwen3-235B,思考模型超越谷歌Gemini-2.5-Flash,成为最强低能耗开源模型之一。","content":"<html><head></head><body><blockquote><p>阿里巴巴开源Qwen3-Next-80B-A3B新架构模型,融合门控DeltaNet和门控注意力的混合架构,训练成本较Qwen3-32B暴降90%,推理效率提升10倍,在超长文本32K以上场景表现尤佳。性能上,指令微调版本媲美旗舰Qwen3-235B,思考模型超越谷歌Gemini-2.5-Flash,成为最强低能耗开源模型之一。</p></blockquote><p style=\"text-align: justify;\">今天凌晨2点,阿里巴巴开源了新架构模型Qwen3-Next-80B-A3B,对混合注意力机制、高稀疏性MoE、训练方法等进行了大幅度创新,迎来了自己的DeepSeek时刻。</p><p style=\"text-align: justify;\">Qwen3-Next是一个混合专家模型总参数800亿,仅激活30亿,训练成本较Qwen3-32B暴降90%,推理效率却提升10倍,尤其是在超长文本32K以上的提示场景中。</p><p style=\"text-align: justify;\">性能方面,Qwen3-Next的指令微调模型在推理与长上下文任务中,可媲美阿里的旗舰模型Qwen3-235B;思考模型则超过了谷歌最新的Gemini-2.5-Flash思考模型,成为目前最强低能耗开源模型之一。</p><p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/3e1195a7f92abd4c7b2f8aecce07bc5a\" tg-width=\"865\" tg-height=\"316\"/></p><p style=\"text-align: justify;\">在线体验:https://chat.qwen.ai/</p><p style=\"text-align: justify;\">开源地址:https://huggingface.co/collections/Qwen/qwen3-next-68c25fd6838e585db8eeea9d</p><p style=\"text-align: justify;\">https://modelscope.cn/collections/Qwen3-Next-c314f23bd0264a</p><p style=\"text-align: justify;\">阿里API:https://www.alibabacloud.com/help/en/model-studio/models#c5414da58bjgj</p><p style=\"text-align: justify;\">网友对阿里新模型的架构非常赞赏,表示,半年前我才刚跟联合创始人聊过类似这样的架构!当时好像把它叫做 “动态权重注意力” 之类的,具体名字记不太清了。这设计真的太出色了!</p><p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/a2c0dde2edca64e3ff02338d600871b8\" tg-width=\"865\" tg-height=\"252\"/></p><p style=\"text-align: justify;\">昨天我测试了好几款模型:思维模式下的 ChatGPT-5、Claude-4,还有专家模式下的 Grok-4。刚刚又测了Qwen3 Next。在所有这些模型里,只有你们这款模型第一次尝试就给了我正确答案。真的太出色了!</p><p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/01c463b2636d8b94e644c8dcc5543634\" tg-width=\"865\" tg-height=\"230\"/></p><p style=\"text-align: justify;\">未来以来,这个模型击败了谷歌的Gemini-2.5-Flash。</p><p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/82eb25aba26a0f3f56c713edaedf46c0\" tg-width=\"865\" tg-height=\"206\"/></p><p style=\"text-align: justify;\">在这里看到 DeltaNet的应用,真的有点让人惊喜!我很好奇,如果换成模型架构发现的AlphaGo 时刻这篇论文中提出的模型架构,这款模型的性能会发生怎样的变化?</p><p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/8716669e82ece67859edf9e0e0b60168\" tg-width=\"865\" tg-height=\"239\"/></p><p style=\"text-align: justify;\">800 亿参数、超高稀疏性再加上多token预测,这配置太惊艳了!要是你的 GPU 有足够显存,用它跑起来速度绝对飞快。</p><p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/6b48dd54ac6114016d90c344e4463f80\" tg-width=\"865\" tg-height=\"176\"/></p><p style=\"text-align: justify;\">基本上老外对阿里的创新模型非常满意,赞美超多。</p><p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/297670e7bad94b6a4eff5411fb13eef2\" tg-width=\"865\" tg-height=\"996\"/></p><p style=\"text-align: justify;\">Qwen3-Next架构简单介绍</p><p style=\"text-align: justify;\">阿里认为上下文长度扩展与总参数扩展是大模型未来发展的两大核心趋势,为在长上下文和大参数场景下进一步提升训练与推理效率,他们设计了全新的模型架构Qwen3-Next。</p><p style=\"text-align: justify;\">相较于Qwen3的MoE结构,Qwen3-Next进行了多项关键改进,包括混合注意力机制、高稀疏性MoE结构、利于训练稳定性的优化手段,以及可实现更快推理的多token预测机制。</p><p style=\"text-align: justify;\">在核心特性方面,Qwen3-Next采用门控DeltaNet+门控注意力的混合创新架构。线性注意力虽能打破标准注意力的二次复杂度,更适合长上下文处理,但仅用线性注意力或标准注意力均有局限。</p><p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/caf800847584cbb841385701513546a8\" tg-width=\"865\" tg-height=\"669\"/></p><p style=\"text-align: justify;\">线性注意力速度快但召回能力弱,标准注意力推理时成本高、速度慢。经系统实验验证,门控DeltaNet的上下文学习能力优于滑动窗口注意力、Mamba2等常用方法,将其与标准注意力按3:1比例,75%层用门控DeltaNet,25%层保留标准注意力结合,模型性能持续超越单一架构,实现性能与效率的双重提升。</p><p style=\"text-align: justify;\">标准注意力层还进行了多项增强,如采用此前研究中的输出门控机制以减少注意力低秩问题、将每个注意力头的维度从128提升至256、仅对前25%位置维度应用旋转位置编码以改善长序列外推能力。</p><p style=\"text-align: justify;\">稀疏性设计上,Qwen3-Next采用超高稀疏性MoE结构,800亿总参数在每步推理中仅激活约30亿,占比3.7%。实验表明,在全局负载均衡的前提下,固定激活专家数量并增加专家总参数,能稳步降低训练损失。与Qwen3的MoE相比,Qwen3-Next将总专家数扩展至512个,结合10个路由专家+1个共享专家的设计,在不影响性能的同时最大化资源利用率。</p><p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/77da51d760e41d13e613c24772e61e65\" tg-width=\"865\" tg-height=\"922\"/></p><p style=\"text-align: justify;\">训练稳定性优化方面,注意力输出门控机制有效解决了注意力Sink、大规模激活等问题,保障模型数值稳定性;针对Qwen3中QK-Norm存在的部分层归一化权重异常增大问题,Qwen3-Next采用零中心RMSNorm,并对归一化权重施加权重衰减以防止无界增长;初始化时对MoE路由器参数进行归一化,确保训练初期每个专家都能被无偏选择,减少随机初始化带来的噪声。这些设计提升了小规模实验的可靠性,保障大规模训练平稳进行。</p><p style=\"text-align: justify;\">多token预测机制也是Qwen3-Next的亮点,其原生引入的多token预测(MTP)机制,不仅为投机解码提供高接受率的MTP模块,还能提升模型整体性能,同时针对MTP的多步推理性能进行优化,通过保持训练与推理一致性的多步训练,进一步提高实际场景中投机解码的接受率。</p><p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/5d434a1876d13581c15d4658a7125a82\" tg-width=\"865\" tg-height=\"519\"/></p><p style=\"text-align: justify;\">预训练阶段,Qwen3-Next展现出卓越的效率。其训练数据来自Qwen3的36T token预训练语料中均匀采样的15T token子集,GPU时长不足Qwen3-30-3B的80%,计算成本仅为Qwen3-32B的9.3%,却能实现更优性能。推理速度上,填充阶段4K上下文长度时吞吐量接近Qwen3-32B的7倍,32K以上时超10倍;</p><p style=\"text-align: justify;\">解码阶段4K上下文长度时吞吐量接近Qwen3-32B的4倍,32K以上时仍保持超10倍的速度优势。性能表现上,Qwen3-Next-80B-A3B-Base仅激活Qwen3-32B-Base非嵌入参数的1/10,却在多数基准测试中性能更优,且显著超过Qwen3-30B-A3B。</p><p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/5c247e4f3ea45e8d086a2a4d3ee5e149\" tg-width=\"1024\" tg-height=\"632\"/></p><p style=\"text-align: justify;\">后训练阶段的性能同样亮眼。指令模型Qwen3-Next-80B-A3B-Instruct大幅超越Qwen3-30B-A3B-Instruct-2507和Qwen3-32B-Non-thinking,性能接近旗舰模型Qwen3-235B-A22B-Instruct-2507;在RULER基准测试中,该模型在各长度下均优于注意力层更多的Qwen3-30B-A3B-Instruct-2507,且在256K上下文内击败总层数更多的Qwen3-235B-A22B-Instruct-2507,印证了混合架构在长上下文任务中的优势。</p><p style=\"text-align: justify;\">推理模型Qwen3-Next-80B-A3B-Thinking性能超过Qwen3-30B-A3B-Thinking-2507、Qwen3-32B-Thinking等更高成本模型,多个基准测试击败Gemini-2.5-Flash-Thinking,关键指标接近Qwen3-235B-A22B-Thinking-2507。</p></body></html>","source":"lsy1680749794970","collect":0,"html":"<!DOCTYPE html>\n<html>\n<head>\n<meta http-equiv=\"Content-Type\" content=\"text/html; charset=utf-8\" />\n<meta name=\"viewport\" content=\"width=device-width,initial-scale=1.0,minimum-scale=1.0,maximum-scale=1.0,user-scalable=no\"/>\n<meta name=\"format-detection\" content=\"telephone=no,email=no,address=no\" />\n<title>阿里DeepSeek时刻!开源新架构模型:推理快10倍、成本暴降90%</title>\n<style type=\"text/css\">\na,abbr,acronym,address,applet,article,aside,audio,b,big,blockquote,body,canvas,caption,center,cite,code,dd,del,details,dfn,div,dl,dt,\nem,embed,fieldset,figcaption,figure,footer,form,h1,h2,h3,h4,h5,h6,header,hgroup,html,i,iframe,img,ins,kbd,label,legend,li,mark,menu,nav,\nobject,ol,output,p,pre,q,ruby,s,samp,section,small,span,strike,strong,sub,summary,sup,table,tbody,td,tfoot,th,thead,time,tr,tt,u,ul,var,video{ font:inherit;margin:0;padding:0;vertical-align:baseline;border:0 }\nbody{ font-size:16px; line-height:1.5; color:#999; background:transparent; }\n.wrapper{ overflow:hidden;word-break:break-all;padding:10px; }\nh1,h2{ font-weight:normal; line-height:1.35; margin-bottom:.6em; }\nh3,h4,h5,h6{ line-height:1.35; margin-bottom:1em; }\nh1{ font-size:24px; }\nh2{ font-size:20px; }\nh3{ font-size:18px; }\nh4{ font-size:16px; }\nh5{ font-size:14px; }\nh6{ font-size:12px; }\np,ul,ol,blockquote,dl,table{ margin:1.2em 0; }\nul,ol{ margin-left:2em; }\nul{ list-style:disc; }\nol{ list-style:decimal; }\nli,li p{ margin:10px 0;}\nimg{ max-width:100%;display:block;margin:0 auto 1em; }\nblockquote{ color:#B5B2B1; border-left:3px solid #aaa; padding:1em; }\nstrong,b{font-weight:bold;}\nem,i{font-style:italic;}\ntable{ width:100%;border-collapse:collapse;border-spacing:1px;margin:1em 0;font-size:.9em; }\nth,td{ padding:5px;text-align:left;border:1px solid #aaa; }\nth{ font-weight:bold;background:#5d5d5d; }\n.symbol-link{font-weight:bold;}\n/* header{ border-bottom:1px solid #494756; } */\n.title{ margin:0 0 8px;line-height:1.3;color:#ddd; }\n.meta {color:#5e5c6d;font-size:13px;margin:0 0 .5em; }\na{text-decoration:none; color:#2a4b87;}\n.meta .head { display: inline-block; overflow: hidden}\n.head .h-thumb { width: 30px; height: 30px; margin: 0; padding: 0; border-radius: 50%; float: left;}\n.head .h-content { margin: 0; padding: 0 0 0 9px; float: left;}\n.head .h-name {font-size: 13px; color: #eee; margin: 0;}\n.head .h-time {font-size: 11px; color: #7E829C; margin: 0;line-height: 11px;}\n.small {font-size: 12.5px; display: inline-block; transform: scale(0.9); -webkit-transform: scale(0.9); transform-origin: left; -webkit-transform-origin: left;}\n.smaller {font-size: 12.5px; display: inline-block; transform: scale(0.8); -webkit-transform: scale(0.8); transform-origin: left; -webkit-transform-origin: left;}\n.bt-text {font-size: 12px;margin: 1.5em 0 0 0}\n.bt-text p {margin: 0}\n</style>\n</head>\n<body>\n<div class=\"wrapper\">\n<header>\n<h2 class=\"title\">\n阿里DeepSeek时刻!开源新架构模型:推理快10倍、成本暴降90%\n</h2>\n\n<h4 class=\"meta\">\n\n\n2025-09-12 08:35 北京时间 <a href=https://mp.weixin.qq.com/s/H2-7AsvsO8HAJzPXBw5Qbg><strong>AIGC开放社区</strong></a>\n\n\n</h4>\n\n</header>\n<article>\n<div>\n<p>阿里巴巴开源Qwen3-Next-80B-A3B新架构模型,融合门控DeltaNet和门控注意力的混合架构,训练成本较Qwen3-32B暴降90%,推理效率提升10倍,在超长文本32K以上场景表现尤佳。性能上,指令微调版本媲美旗舰Qwen3-235B,思考模型超越谷歌Gemini-2.5-Flash,成为最强低能耗开源模型之一。今天凌晨2点,阿里巴巴开源了新架构模型Qwen3-Next-80B-...</p>\n\n<a href=\"https://mp.weixin.qq.com/s/H2-7AsvsO8HAJzPXBw5Qbg\">Web Link</a>\n\n</div>\n\n\n</article>\n</div>\n</body>\n</html>\n","type":0,"thumbnail":"https://static.tigerbbs.com/66a54173e70f4b676e346f3e3ec2f8ea","relate_stocks":{"BABA":"阿里巴巴","09988":"阿里巴巴-W"},"source_url":"https://mp.weixin.qq.com/s/H2-7AsvsO8HAJzPXBw5Qbg","is_english":false,"share_image_url":"https://static.laohu8.com/e9f99090a1c2ed51c021029395664489","article_id":"2566948680","content_text":"阿里巴巴开源Qwen3-Next-80B-A3B新架构模型,融合门控DeltaNet和门控注意力的混合架构,训练成本较Qwen3-32B暴降90%,推理效率提升10倍,在超长文本32K以上场景表现尤佳。性能上,指令微调版本媲美旗舰Qwen3-235B,思考模型超越谷歌Gemini-2.5-Flash,成为最强低能耗开源模型之一。今天凌晨2点,阿里巴巴开源了新架构模型Qwen3-Next-80B-A3B,对混合注意力机制、高稀疏性MoE、训练方法等进行了大幅度创新,迎来了自己的DeepSeek时刻。Qwen3-Next是一个混合专家模型总参数800亿,仅激活30亿,训练成本较Qwen3-32B暴降90%,推理效率却提升10倍,尤其是在超长文本32K以上的提示场景中。性能方面,Qwen3-Next的指令微调模型在推理与长上下文任务中,可媲美阿里的旗舰模型Qwen3-235B;思考模型则超过了谷歌最新的Gemini-2.5-Flash思考模型,成为目前最强低能耗开源模型之一。在线体验:https://chat.qwen.ai/开源地址:https://huggingface.co/collections/Qwen/qwen3-next-68c25fd6838e585db8eeea9dhttps://modelscope.cn/collections/Qwen3-Next-c314f23bd0264a阿里API:https://www.alibabacloud.com/help/en/model-studio/models#c5414da58bjgj网友对阿里新模型的架构非常赞赏,表示,半年前我才刚跟联合创始人聊过类似这样的架构!当时好像把它叫做 “动态权重注意力” 之类的,具体名字记不太清了。这设计真的太出色了!昨天我测试了好几款模型:思维模式下的 ChatGPT-5、Claude-4,还有专家模式下的 Grok-4。刚刚又测了Qwen3 Next。在所有这些模型里,只有你们这款模型第一次尝试就给了我正确答案。真的太出色了!未来以来,这个模型击败了谷歌的Gemini-2.5-Flash。在这里看到 DeltaNet的应用,真的有点让人惊喜!我很好奇,如果换成模型架构发现的AlphaGo 时刻这篇论文中提出的模型架构,这款模型的性能会发生怎样的变化?800 亿参数、超高稀疏性再加上多token预测,这配置太惊艳了!要是你的 GPU 有足够显存,用它跑起来速度绝对飞快。基本上老外对阿里的创新模型非常满意,赞美超多。Qwen3-Next架构简单介绍阿里认为上下文长度扩展与总参数扩展是大模型未来发展的两大核心趋势,为在长上下文和大参数场景下进一步提升训练与推理效率,他们设计了全新的模型架构Qwen3-Next。相较于Qwen3的MoE结构,Qwen3-Next进行了多项关键改进,包括混合注意力机制、高稀疏性MoE结构、利于训练稳定性的优化手段,以及可实现更快推理的多token预测机制。在核心特性方面,Qwen3-Next采用门控DeltaNet+门控注意力的混合创新架构。线性注意力虽能打破标准注意力的二次复杂度,更适合长上下文处理,但仅用线性注意力或标准注意力均有局限。线性注意力速度快但召回能力弱,标准注意力推理时成本高、速度慢。经系统实验验证,门控DeltaNet的上下文学习能力优于滑动窗口注意力、Mamba2等常用方法,将其与标准注意力按3:1比例,75%层用门控DeltaNet,25%层保留标准注意力结合,模型性能持续超越单一架构,实现性能与效率的双重提升。标准注意力层还进行了多项增强,如采用此前研究中的输出门控机制以减少注意力低秩问题、将每个注意力头的维度从128提升至256、仅对前25%位置维度应用旋转位置编码以改善长序列外推能力。稀疏性设计上,Qwen3-Next采用超高稀疏性MoE结构,800亿总参数在每步推理中仅激活约30亿,占比3.7%。实验表明,在全局负载均衡的前提下,固定激活专家数量并增加专家总参数,能稳步降低训练损失。与Qwen3的MoE相比,Qwen3-Next将总专家数扩展至512个,结合10个路由专家+1个共享专家的设计,在不影响性能的同时最大化资源利用率。训练稳定性优化方面,注意力输出门控机制有效解决了注意力Sink、大规模激活等问题,保障模型数值稳定性;针对Qwen3中QK-Norm存在的部分层归一化权重异常增大问题,Qwen3-Next采用零中心RMSNorm,并对归一化权重施加权重衰减以防止无界增长;初始化时对MoE路由器参数进行归一化,确保训练初期每个专家都能被无偏选择,减少随机初始化带来的噪声。这些设计提升了小规模实验的可靠性,保障大规模训练平稳进行。多token预测机制也是Qwen3-Next的亮点,其原生引入的多token预测(MTP)机制,不仅为投机解码提供高接受率的MTP模块,还能提升模型整体性能,同时针对MTP的多步推理性能进行优化,通过保持训练与推理一致性的多步训练,进一步提高实际场景中投机解码的接受率。预训练阶段,Qwen3-Next展现出卓越的效率。其训练数据来自Qwen3的36T token预训练语料中均匀采样的15T token子集,GPU时长不足Qwen3-30-3B的80%,计算成本仅为Qwen3-32B的9.3%,却能实现更优性能。推理速度上,填充阶段4K上下文长度时吞吐量接近Qwen3-32B的7倍,32K以上时超10倍;解码阶段4K上下文长度时吞吐量接近Qwen3-32B的4倍,32K以上时仍保持超10倍的速度优势。性能表现上,Qwen3-Next-80B-A3B-Base仅激活Qwen3-32B-Base非嵌入参数的1/10,却在多数基准测试中性能更优,且显著超过Qwen3-30B-A3B。后训练阶段的性能同样亮眼。指令模型Qwen3-Next-80B-A3B-Instruct大幅超越Qwen3-30B-A3B-Instruct-2507和Qwen3-32B-Non-thinking,性能接近旗舰模型Qwen3-235B-A22B-Instruct-2507;在RULER基准测试中,该模型在各长度下均优于注意力层更多的Qwen3-30B-A3B-Instruct-2507,且在256K上下文内击败总层数更多的Qwen3-235B-A22B-Instruct-2507,印证了混合架构在长上下文任务中的优势。推理模型Qwen3-Next-80B-A3B-Thinking性能超过Qwen3-30B-A3B-Thinking-2507、Qwen3-32B-Thinking等更高成本模型,多个基准测试击败Gemini-2.5-Flash-Thinking,关键指标接近Qwen3-235B-A22B-Thinking-2507。","news_type":1,"symbols_score_info":{"09988":1.1,"BABA":1}},"isVote":1,"tweetType":1,"viewCount":185,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":0,"langContent":"EN","totalScore":0},{"id":469317663191080,"gmtCreate":1755610564313,"gmtModify":1755613695043,"author":{"id":"3494390725694021","authorId":"3494390725694021","name":"charmby","avatar":"https://static.tigerbbs.com/c0555fde187648aa2f83f9fe88db109b","crmLevel":9,"crmLevelSwitch":1,"followedFlag":false,"authorIdStr":"3494390725694021","idStr":"3494390725694021"},"themes":[],"htmlText":"<a href=\"https://laohu8.com/S/MSTR\">$Strategy(MSTR)$ </a> 我靠 这货期权没救了 跌一个点期权就跌50个点 涨一个点 期权涨幅为0。。这样的谁在期权上赚过钱啊","listText":"<a href=\"https://laohu8.com/S/MSTR\">$Strategy(MSTR)$ </a> 我靠 这货期权没救了 跌一个点期权就跌50个点 涨一个点 期权涨幅为0。。这样的谁在期权上赚过钱啊","text":"$Strategy(MSTR)$ 我靠 这货期权没救了 跌一个点期权就跌50个点 涨一个点 期权涨幅为0。。这样的谁在期权上赚过钱啊","images":[],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":2,"commentSize":1,"repostSize":2,"link":"https://laohu8.com/post/469317663191080","isVote":1,"tweetType":1,"viewCount":2050,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":0,"langContent":"CN","totalScore":0},{"id":468987766067672,"gmtCreate":1755530220878,"gmtModify":1755530222101,"author":{"id":"3494390725694021","authorId":"3494390725694021","name":"charmby","avatar":"https://static.tigerbbs.com/c0555fde187648aa2f83f9fe88db109b","crmLevel":9,"crmLevelSwitch":1,"followedFlag":false,"authorIdStr":"3494390725694021","idStr":"3494390725694021"},"themes":[],"htmlText":"<a href=\"https://laohu8.com/S/MSTR\">$Strategy(MSTR)$ </a><v-v data-views=\"1\"></v-v> 这个怎么回事。。。拉起来一半到0.5 居然call创新低????","listText":"<a href=\"https://laohu8.com/S/MSTR\">$Strategy(MSTR)$ </a><v-v data-views=\"1\"></v-v> 这个怎么回事。。。拉起来一半到0.5 居然call创新低????","text":"$Strategy(MSTR)$ 这个怎么回事。。。拉起来一半到0.5 居然call创新低????","images":[],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":0,"commentSize":0,"repostSize":0,"link":"https://laohu8.com/post/468987766067672","isVote":1,"tweetType":1,"viewCount":602,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":0,"langContent":"CN","totalScore":0},{"id":455489876877824,"gmtCreate":1752240809248,"gmtModify":1752501114227,"author":{"id":"3494390725694021","authorId":"3494390725694021","name":"charmby","avatar":"https://static.tigerbbs.com/c0555fde187648aa2f83f9fe88db109b","crmLevel":9,"crmLevelSwitch":1,"followedFlag":false,"authorIdStr":"3494390725694021","idStr":"3494390725694021"},"themes":[],"htmlText":"我开仓了55手<a href=\"https://laohu8.com/OPT/NVDA 20250718 160.0 PUT\">$NVDA 20250718 160.0 PUT$ </a>,来看看我最新分享的订单!","listText":"我开仓了55手<a href=\"https://laohu8.com/OPT/NVDA 20250718 160.0 PUT\">$NVDA 20250718 160.0 PUT$ </a>,来看看我最新分享的订单!","text":"我开仓了55手$NVDA 20250718 160.0 PUT$ ,来看看我最新分享的订单!","images":[],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":0,"commentSize":0,"repostSize":0,"link":"https://laohu8.com/post/455489876877824","isVote":1,"tweetType":1,"viewCount":668,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":0,"langContent":"CN","totalScore":0},{"id":452010886472360,"gmtCreate":1751380207825,"gmtModify":1751381558648,"author":{"id":"3494390725694021","authorId":"3494390725694021","name":"charmby","avatar":"https://static.tigerbbs.com/c0555fde187648aa2f83f9fe88db109b","crmLevel":9,"crmLevelSwitch":1,"followedFlag":false,"authorIdStr":"3494390725694021","idStr":"3494390725694021"},"themes":[],"htmlText":"我平仓了63手<a href=\"https://laohu8.com/OPT/HOOD 20250711 87.0 PUT\">$HOOD 20250711 87.0 PUT$ </a>,要多倒霉有多倒霉[流泪] [流泪] [流泪] aoao","listText":"我平仓了63手<a href=\"https://laohu8.com/OPT/HOOD 20250711 87.0 PUT\">$HOOD 20250711 87.0 PUT$ </a>,要多倒霉有多倒霉[流泪] [流泪] [流泪] aoao","text":"我平仓了63手$HOOD 20250711 87.0 PUT$ ,要多倒霉有多倒霉[流泪] [流泪] [流泪] aoao","images":[],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":0,"commentSize":0,"repostSize":0,"link":"https://laohu8.com/post/452010886472360","isVote":1,"tweetType":1,"viewCount":759,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":0,"langContent":"CN","totalScore":0},{"id":452012043629136,"gmtCreate":1751380045262,"gmtModify":1751380243890,"author":{"id":"3494390725694021","authorId":"3494390725694021","name":"charmby","avatar":"https://static.tigerbbs.com/c0555fde187648aa2f83f9fe88db109b","crmLevel":9,"crmLevelSwitch":1,"followedFlag":false,"authorIdStr":"3494390725694021","idStr":"3494390725694021"},"themes":[],"htmlText":"我平仓了55手<a href=\"https://laohu8.com/OPT/NVDA 20250711 157.5 PUT\">$NVDA 20250711 157.5 PUT$ </a>,每次都被另一只坑 导致每次节奏都乱 tnd","listText":"我平仓了55手<a href=\"https://laohu8.com/OPT/NVDA 20250711 157.5 PUT\">$NVDA 20250711 157.5 PUT$ </a>,每次都被另一只坑 导致每次节奏都乱 tnd","text":"我平仓了55手$NVDA 20250711 157.5 PUT$ ,每次都被另一只坑 导致每次节奏都乱 tnd","images":[],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":0,"commentSize":0,"repostSize":0,"link":"https://laohu8.com/post/452012043629136","isVote":1,"tweetType":1,"viewCount":909,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":0,"langContent":"CN","totalScore":0},{"id":451514553749504,"gmtCreate":1751259292054,"gmtModify":1751260196777,"author":{"id":"3494390725694021","authorId":"3494390725694021","name":"charmby","avatar":"https://static.tigerbbs.com/c0555fde187648aa2f83f9fe88db109b","crmLevel":9,"crmLevelSwitch":1,"followedFlag":false,"authorIdStr":"3494390725694021","idStr":"3494390725694021"},"themes":[],"htmlText":"原本备受看好的投资标的(美元、美股等)表现惨淡, 来搞笑了 别人历史新高","listText":"原本备受看好的投资标的(美元、美股等)表现惨淡, 来搞笑了 别人历史新高","text":"原本备受看好的投资标的(美元、美股等)表现惨淡, 来搞笑了 别人历史新高","images":[],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":0,"commentSize":0,"repostSize":0,"link":"https://laohu8.com/post/451514553749504","repostId":"2547053912","repostType":4,"repost":{"id":"2547053912","kind":"news","pubTimestamp":1751257194,"share":"https://www.laohu8.com/m/news/2547053912?lang=&edition=full","pubTime":"2025-06-30 12:19","market":"us","language":"zh","title":"半年过去了,华尔街的“脸都被打肿了”","url":"https://stock-news.laohu8.com/highlight/detail?id=2547053912","media":"华尔街见闻","summary":"特朗普关税政策和地缘政治冲突彻底颠覆了华尔街年初预测。美元遭遇2005年来最差年初表现,标普500经历惊人暴跌和闪电式反弹。欧洲股市从投资洼地变为必备资产,基准指数跑赢标普500达16个百分点,新兴市场打破连年跑输魔咒,货币普遍兑美元走强,股市为股东创造1.8万亿美元财富增长。","content":"<html><head></head><body><p>特朗普关税政策和地缘政治冲突彻底颠覆了华尔街年初的预测,原本备受看好的投资标的(美元、美股等)表现惨淡,意想不到的赢家(欧洲股市、新兴市场)却脱颖而出。</p><p>自2025年初华尔街发布预测以来的六个月里,特朗普政策不确定性和世界冲突彻底打破了市场对美国资产和经济实力及主导地位的假设。</p><p>美元遭遇自2005年以来最糟糕的年初表现,标普500指数经历了惊人的暴跌和闪电式的反弹。与此同时,<strong>欧洲股市从投资洼地摇身一变成为投资者必备资产</strong>。</p><p>新兴市场终于迎来复苏,人工智能公司蓬勃发展,货币普遍兑美元走强,<strong>2025年上半年为股东创造了1.8万亿美元的财富增长</strong>。</p><p><a href=\"https://laohu8.com/S/GS\">高盛</a>资产管理固定收益宏观策略主管Simon Dangoor表示,<strong>过去六个月市场发生了"非常重大的演变",年初基于中期趋势制定的任何主题都受到了考验</strong>。</p><h2 id=\"id_648985330\">美元霸权地位动摇</h2><p>特朗普的低税收、高关税政策原本被预期将推高通胀并降低美联储降息可能性,这些因素本应推动美元在2025年保持强势。然而现实截然相反,<a href=\"https://laohu8.com/S/USDindex.FOREX\">美元指数</a>录得自2005年以来最差的年初表现。</p><p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/054a27eefbc32ffa78421356725b5acf\" tg-width=\"1024\" tg-height=\"673\"/></p><p>特朗普4月初公布了所谓的"对等关税”范围广泛且极具惩罚性,引发了对美国经济衰退的担忧,并催生了特朗普试图通过美元贬值来提振国内制造业的猜测。</p><p><a href=\"https://laohu8.com/S/0J6Y.UK\">法国兴业银行</a>、<a href=\"https://laohu8.com/S/MS\">摩根士丹利</a>和<a href=\"https://laohu8.com/S/JPM\">摩根大通</a>此前都未预期美元在上半年出现转折。</p><p>据报道,由Meera Chandan领导的摩根大通团队最新表示,美元与利率和股票联系的减弱可能是结构性弱点的信号,预计美元强势指标到年底将再下跌2%。</p><h2 id=\"id_4117194003\">美股大起大落后重获青睐</h2><p>投资者年初对美股配置创纪录新高,受强劲经济和人工智能押注鼓舞。然而这种乐观情绪在几个月内几乎完全消失,先是中国初创公司DeepSeek挑战美国在AI竞赛中的主导地位,随后特朗普关税引发经济衰退担忧。</p><p>科技股集中的纳斯达克100指数在2月峰值至4月低点间蒸发了近7万亿美元市值。<a href=\"https://laohu8.com/S/BAC\">美国银行</a>基金经理调查显示,3月份美股敞口出现有史以来最大降幅。</p><p>但特朗普4月下旬暂停部分高关税的决定成为转折点。标普500指数创下历史新高,经济数据显示增长稳健,科技股再度受到追捧。</p><p>State Street Global Markets高级多资产策略师Marija Veitmane表示:</p><blockquote><p>"我对美股一如既往看涨,它们仍提供最佳盈利前景,增长最快且最具可预测性。"</p></blockquote><h2 id=\"id_3181397212\">亚洲货币强势反弹</h2><p>在全球其他央行纷纷降息之际,日本央行准备加息,交易员们在2025年伊始对日元反弹充满信心。在今年年初的时候,<a href=\"https://laohu8.com/S/USDJPY.FOREX\">美元/日元</a>交投于159水平附近。</p><p>六个月之后,摩根大通资产管理公司和Brandywine全球投资管理公司被证明是正确的,今年以来,美元兑日元跌近9%,目前<a href=\"https://laohu8.com/S/002200\">交投</a>于145左右。</p><p>在特朗普关税引发的市场混乱中,对避险资产的需求激增,日元在4月得到了显著提振。美元/日元跌4.6%,最低跌至139.89。</p><p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/9beb5102c8637f7f58f711a0db097ff1\" tg-width=\"1024\" tg-height=\"675\"/></p><p>Jupiter Asset Management的纳什(Mark Nash)在今年1月就开始为日元升值做准备。他预计到年底日元将升至1美元兑120元人民币,较目前水平上涨约17%。</p><p>与此同时,美国贸易关税原本被预期将冲击人民币,但迄今为止美元自身的急剧抛售颠覆了这一预测。</p><p>野村在12月预测人民币离岸汇率将在5月份跌至每美元7.6,摩根大通预期二季度为7.5。相反,人民币今年飙升1.8%,上周四一度触及7.1565,为七个月高点。</p><h2 id=\"id_2466637238\">全球债券分化加剧</h2><p>NatWest Markets固定收益交易全球主管Jared Noering表示,在动荡中,许多投资者对一笔“救了他们一命”的交易心存感激。</p><p>短期政府债券在央行因通胀进一步缓解而降息的推动下表现良好。相比之下,长期债券因政府承担更多债务以填补日益加深的财政赤字并增加公共支出而面临压力。</p><p>围绕这种分歧构建的押注基本上在全球各地都有出现,包括在美国,市场仍对政府的税收和支出计划感到不安。衡量较长期美国国债所谓期限溢价的指标大幅飙升,表明买家正要求为猖獗的借款提供更高的补偿。</p><p><a href=\"https://laohu8.com/S/601099\">太平洋</a>投资管理公司和Allspring Global Investments正确预测了全球债券市场短期和长期收益率的差异。<a href=\"https://laohu8.com/S/BLK\">贝莱德</a>减持长期美国国债也是正确的。</p><h2 id=\"id_3130662695\">欧股和新兴市场成最大赢家</h2><p>年初很难找到欧洲股票的拥趸,更别说押注其将跑赢美股的投行了。但是六个月后,对经济疲软和关税威胁的担忧被德国释放数千亿欧元国防支出的计划所抵消,</p><p>截至6月27日,基准斯托克600指数以美元计价跑赢标普500指数16个百分点,创2016年以来最佳相对表现。<a href=\"https://laohu8.com/S/EURUSD.FOREX\">欧元/美元</a>飙升至1.17美元,打破了年初与美元平价的普遍预测。</p><p>花旗集团欧洲和全球股票策略主管Beata Manthey是去年底支持欧洲股票的少数声音之一。摩根大通和高盛的目标被证明过于谨慎。高盛首席全球股票策略师Peter Oppenheimer表示,</p><blockquote><p>“情况已大不相同,非常激进的关税不太可能完全实施。"</p></blockquote><p>新兴市场方面,自2017年以来每年都跑输美股的魔咒终于被打破。来自中国、韩国等人工智能公司繁荣发展,加上货币兑美元普遍走强,新兴市场2025年已为股东增加1.8万亿美元财富,市值达到创纪录的29万亿美元。</p><p>InTouch Capital Markets策略师Bernd Berg表示,得益于温和通胀和可观增长率,预计资金流入将持续。Berg表示:</p><blockquote><p>"地缘政治紧张局势并未破坏这轮涨势。"</p></blockquote><p>风险提示及免责条款</p><p> 市场有风险,投资需谨慎。本文不构成个人投资建议,也未考虑到个别用户特殊的投资目标、财务状况或需要。用户应考虑本文中的任何意见、观点或结论是否符合其特定状况。据此投资,责任自负。</p></body></html>","source":"wallstreetcn_hot_news","collect":0,"html":"<!DOCTYPE html>\n<html>\n<head>\n<meta http-equiv=\"Content-Type\" content=\"text/html; charset=utf-8\" />\n<meta name=\"viewport\" content=\"width=device-width,initial-scale=1.0,minimum-scale=1.0,maximum-scale=1.0,user-scalable=no\"/>\n<meta name=\"format-detection\" content=\"telephone=no,email=no,address=no\" />\n<title>半年过去了,华尔街的“脸都被打肿了”</title>\n<style type=\"text/css\">\na,abbr,acronym,address,applet,article,aside,audio,b,big,blockquote,body,canvas,caption,center,cite,code,dd,del,details,dfn,div,dl,dt,\nem,embed,fieldset,figcaption,figure,footer,form,h1,h2,h3,h4,h5,h6,header,hgroup,html,i,iframe,img,ins,kbd,label,legend,li,mark,menu,nav,\nobject,ol,output,p,pre,q,ruby,s,samp,section,small,span,strike,strong,sub,summary,sup,table,tbody,td,tfoot,th,thead,time,tr,tt,u,ul,var,video{ font:inherit;margin:0;padding:0;vertical-align:baseline;border:0 }\nbody{ font-size:16px; line-height:1.5; color:#999; background:transparent; }\n.wrapper{ overflow:hidden;word-break:break-all;padding:10px; }\nh1,h2{ font-weight:normal; line-height:1.35; margin-bottom:.6em; }\nh3,h4,h5,h6{ line-height:1.35; margin-bottom:1em; }\nh1{ font-size:24px; }\nh2{ font-size:20px; }\nh3{ font-size:18px; }\nh4{ font-size:16px; }\nh5{ font-size:14px; }\nh6{ font-size:12px; }\np,ul,ol,blockquote,dl,table{ margin:1.2em 0; }\nul,ol{ margin-left:2em; }\nul{ list-style:disc; }\nol{ list-style:decimal; }\nli,li p{ margin:10px 0;}\nimg{ max-width:100%;display:block;margin:0 auto 1em; }\nblockquote{ color:#B5B2B1; border-left:3px solid #aaa; padding:1em; }\nstrong,b{font-weight:bold;}\nem,i{font-style:italic;}\ntable{ width:100%;border-collapse:collapse;border-spacing:1px;margin:1em 0;font-size:.9em; }\nth,td{ padding:5px;text-align:left;border:1px solid #aaa; }\nth{ font-weight:bold;background:#5d5d5d; }\n.symbol-link{font-weight:bold;}\n/* header{ border-bottom:1px solid #494756; } */\n.title{ margin:0 0 8px;line-height:1.3;color:#ddd; }\n.meta {color:#5e5c6d;font-size:13px;margin:0 0 .5em; }\na{text-decoration:none; color:#2a4b87;}\n.meta .head { display: inline-block; overflow: hidden}\n.head .h-thumb { width: 30px; height: 30px; margin: 0; padding: 0; border-radius: 50%; float: left;}\n.head .h-content { margin: 0; padding: 0 0 0 9px; float: left;}\n.head .h-name {font-size: 13px; color: #eee; margin: 0;}\n.head .h-time {font-size: 11px; color: #7E829C; margin: 0;line-height: 11px;}\n.small {font-size: 12.5px; display: inline-block; transform: scale(0.9); -webkit-transform: scale(0.9); transform-origin: left; -webkit-transform-origin: left;}\n.smaller {font-size: 12.5px; display: inline-block; transform: scale(0.8); -webkit-transform: scale(0.8); transform-origin: left; -webkit-transform-origin: left;}\n.bt-text {font-size: 12px;margin: 1.5em 0 0 0}\n.bt-text p {margin: 0}\n</style>\n</head>\n<body>\n<div class=\"wrapper\">\n<header>\n<h2 class=\"title\">\n半年过去了,华尔街的“脸都被打肿了”\n</h2>\n\n<h4 class=\"meta\">\n\n\n2025-06-30 12:19 北京时间 <a href=https://wallstreetcn.com/articles/3750070><strong>华尔街见闻</strong></a>\n\n\n</h4>\n\n</header>\n<article>\n<div>\n<p>特朗普关税政策和地缘政治冲突彻底颠覆了华尔街年初的预测,原本备受看好的投资标的(美元、美股等)表现惨淡,意想不到的赢家(欧洲股市、新兴市场)却脱颖而出。自2025年初华尔街发布预测以来的六个月里,特朗普政策不确定性和世界冲突彻底打破了市场对美国资产和经济实力及主导地位的假设。美元遭遇自2005年以来最糟糕的年初表现,标普500指数经历了惊人的暴跌和闪电式的反弹。与此同时,欧洲股市从投资洼地摇身一变...</p>\n\n<a href=\"https://wallstreetcn.com/articles/3750070\">Web Link</a>\n\n</div>\n\n\n</article>\n</div>\n</body>\n</html>\n","type":0,"thumbnail":"https://static.tigerbbs.com/9d4287e55e5fbfd6c02d7bf89e2405e3","relate_stocks":{"BK4504":"桥水持仓","UPRO":"三倍做多标普500ETF","BK4559":"巴菲特持仓","BK4550":"红杉资本持仓","BK4588":"碎股","OEF":"标普100指数ETF-iShares","SSO":"两倍做多标普500ETF",".SPX":"S&P 500 Index","SPY":"标普500ETF","BK4581":"高盛持仓","SPXU":"三倍做空标普500ETF","SDS":"两倍做空标普500ETF","BK4534":"瑞士信贷持仓","BK4585":"ETF&股票定投概念","OEX":"标普100","SH":"标普500反向ETF","IVV":"标普500指数ETF"},"source_url":"https://wallstreetcn.com/articles/3750070","is_english":false,"share_image_url":"https://static.laohu8.com/e9f99090a1c2ed51c021029395664489","article_id":"2547053912","content_text":"特朗普关税政策和地缘政治冲突彻底颠覆了华尔街年初的预测,原本备受看好的投资标的(美元、美股等)表现惨淡,意想不到的赢家(欧洲股市、新兴市场)却脱颖而出。自2025年初华尔街发布预测以来的六个月里,特朗普政策不确定性和世界冲突彻底打破了市场对美国资产和经济实力及主导地位的假设。美元遭遇自2005年以来最糟糕的年初表现,标普500指数经历了惊人的暴跌和闪电式的反弹。与此同时,欧洲股市从投资洼地摇身一变成为投资者必备资产。新兴市场终于迎来复苏,人工智能公司蓬勃发展,货币普遍兑美元走强,2025年上半年为股东创造了1.8万亿美元的财富增长。高盛资产管理固定收益宏观策略主管Simon Dangoor表示,过去六个月市场发生了\"非常重大的演变\",年初基于中期趋势制定的任何主题都受到了考验。美元霸权地位动摇特朗普的低税收、高关税政策原本被预期将推高通胀并降低美联储降息可能性,这些因素本应推动美元在2025年保持强势。然而现实截然相反,美元指数录得自2005年以来最差的年初表现。特朗普4月初公布了所谓的\"对等关税”范围广泛且极具惩罚性,引发了对美国经济衰退的担忧,并催生了特朗普试图通过美元贬值来提振国内制造业的猜测。法国兴业银行、摩根士丹利和摩根大通此前都未预期美元在上半年出现转折。据报道,由Meera Chandan领导的摩根大通团队最新表示,美元与利率和股票联系的减弱可能是结构性弱点的信号,预计美元强势指标到年底将再下跌2%。美股大起大落后重获青睐投资者年初对美股配置创纪录新高,受强劲经济和人工智能押注鼓舞。然而这种乐观情绪在几个月内几乎完全消失,先是中国初创公司DeepSeek挑战美国在AI竞赛中的主导地位,随后特朗普关税引发经济衰退担忧。科技股集中的纳斯达克100指数在2月峰值至4月低点间蒸发了近7万亿美元市值。美国银行基金经理调查显示,3月份美股敞口出现有史以来最大降幅。但特朗普4月下旬暂停部分高关税的决定成为转折点。标普500指数创下历史新高,经济数据显示增长稳健,科技股再度受到追捧。State Street Global Markets高级多资产策略师Marija Veitmane表示:\"我对美股一如既往看涨,它们仍提供最佳盈利前景,增长最快且最具可预测性。\"亚洲货币强势反弹在全球其他央行纷纷降息之际,日本央行准备加息,交易员们在2025年伊始对日元反弹充满信心。在今年年初的时候,美元/日元交投于159水平附近。六个月之后,摩根大通资产管理公司和Brandywine全球投资管理公司被证明是正确的,今年以来,美元兑日元跌近9%,目前交投于145左右。在特朗普关税引发的市场混乱中,对避险资产的需求激增,日元在4月得到了显著提振。美元/日元跌4.6%,最低跌至139.89。Jupiter Asset Management的纳什(Mark Nash)在今年1月就开始为日元升值做准备。他预计到年底日元将升至1美元兑120元人民币,较目前水平上涨约17%。与此同时,美国贸易关税原本被预期将冲击人民币,但迄今为止美元自身的急剧抛售颠覆了这一预测。野村在12月预测人民币离岸汇率将在5月份跌至每美元7.6,摩根大通预期二季度为7.5。相反,人民币今年飙升1.8%,上周四一度触及7.1565,为七个月高点。全球债券分化加剧NatWest Markets固定收益交易全球主管Jared Noering表示,在动荡中,许多投资者对一笔“救了他们一命”的交易心存感激。短期政府债券在央行因通胀进一步缓解而降息的推动下表现良好。相比之下,长期债券因政府承担更多债务以填补日益加深的财政赤字并增加公共支出而面临压力。围绕这种分歧构建的押注基本上在全球各地都有出现,包括在美国,市场仍对政府的税收和支出计划感到不安。衡量较长期美国国债所谓期限溢价的指标大幅飙升,表明买家正要求为猖獗的借款提供更高的补偿。太平洋投资管理公司和Allspring Global Investments正确预测了全球债券市场短期和长期收益率的差异。贝莱德减持长期美国国债也是正确的。欧股和新兴市场成最大赢家年初很难找到欧洲股票的拥趸,更别说押注其将跑赢美股的投行了。但是六个月后,对经济疲软和关税威胁的担忧被德国释放数千亿欧元国防支出的计划所抵消,截至6月27日,基准斯托克600指数以美元计价跑赢标普500指数16个百分点,创2016年以来最佳相对表现。欧元/美元飙升至1.17美元,打破了年初与美元平价的普遍预测。花旗集团欧洲和全球股票策略主管Beata Manthey是去年底支持欧洲股票的少数声音之一。摩根大通和高盛的目标被证明过于谨慎。高盛首席全球股票策略师Peter Oppenheimer表示,“情况已大不相同,非常激进的关税不太可能完全实施。\"新兴市场方面,自2017年以来每年都跑输美股的魔咒终于被打破。来自中国、韩国等人工智能公司繁荣发展,加上货币兑美元普遍走强,新兴市场2025年已为股东增加1.8万亿美元财富,市值达到创纪录的29万亿美元。InTouch Capital Markets策略师Bernd Berg表示,得益于温和通胀和可观增长率,预计资金流入将持续。Berg表示:\"地缘政治紧张局势并未破坏这轮涨势。\"风险提示及免责条款 市场有风险,投资需谨慎。本文不构成个人投资建议,也未考虑到个别用户特殊的投资目标、财务状况或需要。用户应考虑本文中的任何意见、观点或结论是否符合其特定状况。据此投资,责任自负。","news_type":1,"symbols_score_info":{".SPX":0.6,"ESmain":0.6,"MESmain":0.6,"IVV":0.6,"OEF":0.6,"OEX":0.6,"SDS":0.6,"SH":0.6,"SPXU":0.6,"SPY":0.67,"SSO":0.6,"UPRO":0.6}},"isVote":1,"tweetType":1,"viewCount":858,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":0,"langContent":"CN","totalScore":0},{"id":450627523408208,"gmtCreate":1751042772887,"gmtModify":1751042775460,"author":{"id":"3494390725694021","authorId":"3494390725694021","name":"charmby","avatar":"https://static.tigerbbs.com/c0555fde187648aa2f83f9fe88db109b","crmLevel":9,"crmLevelSwitch":1,"followedFlag":false,"authorIdStr":"3494390725694021","idStr":"3494390725694021"},"themes":[],"htmlText":"我平仓了47手<a href=\"https://laohu8.com/OPT/NVDA 20250711 157.5 PUT\">$NVDA 20250711 157.5 PUT$ </a>,来看看我最新分享的订单!","listText":"我平仓了47手<a href=\"https://laohu8.com/OPT/NVDA 20250711 157.5 PUT\">$NVDA 20250711 157.5 PUT$ </a>,来看看我最新分享的订单!","text":"我平仓了47手$NVDA 20250711 157.5 PUT$ ,来看看我最新分享的订单!","images":[],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":0,"commentSize":0,"repostSize":0,"link":"https://laohu8.com/post/450627523408208","isVote":1,"tweetType":1,"viewCount":1064,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":0,"langContent":"CN","totalScore":0},{"id":448123393728944,"gmtCreate":1750431374112,"gmtModify":1750431647320,"author":{"id":"3494390725694021","authorId":"3494390725694021","name":"charmby","avatar":"https://static.tigerbbs.com/c0555fde187648aa2f83f9fe88db109b","crmLevel":9,"crmLevelSwitch":1,"followedFlag":false,"authorIdStr":"3494390725694021","idStr":"3494390725694021"},"themes":[],"htmlText":"我平仓了108手<a href=\"https://laohu8.com/OPT/NVDA 20250627 140.0 PUT\">$NVDA 20250627 140.0 PUT$ </a>,!有病 开盘诱多 忍不住全割了 fcccccck","listText":"我平仓了108手<a href=\"https://laohu8.com/OPT/NVDA 20250627 140.0 PUT\">$NVDA 20250627 140.0 PUT$ </a>,!有病 开盘诱多 忍不住全割了 fcccccck","text":"我平仓了108手$NVDA 20250627 140.0 PUT$ ,!有病 开盘诱多 忍不住全割了 fcccccck","images":[],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":1,"commentSize":0,"repostSize":0,"link":"https://laohu8.com/post/448123393728944","isVote":1,"tweetType":1,"viewCount":721,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":0,"langContent":"CN","totalScore":0}],"defaultTab":"following","isTTM":false}