阿里要继续大投入淘宝闪购，三年不担心亏损 《晚点 LatePost》独家获悉，阿里核心管理层 2026 年初在一次内部会议上鼓励团队要继续大胆做闪 购，三年内不要有（亏损的）负担。 阿里创始人马云曾在内部表示，"淘宝闪购是集团里程碑意义的战役。" 据我们了解，淘宝闪购在 2026 年全年投入力度会超过前一年，其重点是即时零售。 前述内部交流也明确了淘宝闪购的几个重点： 餐饮外卖业务，要有节制、更有目的。重点是做 30 元以上的高客单价订单，以及多维度提升配送的精 细运营。 即时零售，需要一个一个找细分的品类重点做，优先做速度要求超过传统电商能力、年销售额 100 亿元 以上的生意，比如医药、酒饮、生鲜等。2026 年即时零售的第一优先级是建仓。 到店业务目前规模尚小。阿里管理层的态度是 2026 年继续要做，探索同城、旅游等各种场景对到店业 务的带动。 淘宝闪购原本的计划是在 3 月中旬发起第一场战役，那时候骑手、商家都已从家乡返回，正常经营；到 夏天，会再次发起大战，实现 2026 年市场份额超过美团的目标。 来源：晚点LatePost 文丨管艺雯 新一轮即时零售大战随着 AI 助手混战提前开始了。 但 2 ...

摘要： 千问战役，值得让整个行业思考，人们到底需要怎样的AI？ 这大概是此次AI春节档最让人意外的数字。 2月12日，千问APP发布"春节30亿大免单"第一波活动数据——过去6天，用户说了41亿次"千问帮我"，AI下单量突破1.2亿笔。 就在一天前，QuestMobile一张数据截图在圈里疯传。2月7日，千问DAU达到7352万，已几乎逼近当前排名第一的豆包。要知道，千问App诞生不过区区 80天，就走完了对方三年的路。 这意味着，阿里用一场"30亿请客"，将自己一举推向了AI牌桌的C位。 更让对手难安的或许是另一个维度，千问代表的"AI办事"应用火爆，意味着用户使用互联网的习惯改变和养成正在发生，"千问帮我"逐步演变为了AI时代 的用户新习惯。 这背后，不仅仅是AI技术的突飞猛进，更是商业生态的降维打击。 24小时闪电战 2月6日零点，千问"春节30亿大免单"活动正式上线。 规则简单粗暴：下载App，领25元免单卡，一分钱点奶茶。邀请好友，再得一张。 这种玩法并不新鲜。新鲜的是阿里这次的"狠"——每人最高可得21张卡，合计525块钱，不限品类，不设门槛。当夜，已有不少圈内人士按捺不住尝试， 在半夜就开始了点 ...

ABot-M0模型架构图 ABot-M0 基于全球开源资源，整合超过 600 万条真实操作轨迹，构建了目前规模最大的通用机器人数据集。通过统一动作表示、坐标系与控制频率，并采 用增量式动作建模，实现了跨平台数据融合，支撑了完全基于公开数据的预训练。 算法革新上，ABot-M0 提出了全球首个动作流形学习：有效的机器人动作受限于物理规律、任务目标与环境约束，集中分布在低维结构化的流形上。设计 了 AML（Action Manifold Learning）算法，使模型能够直接预测结构合理、物理可行的动作序列，提升策略的稳定性与解码效率。 为增强空间感知，ABot-M0 引入 3D 感知模块，增强模型对"前后、远近、遮挡"等空间语义的理解，在复杂环境中实现更精准的操作决策。 在 Libero、Libero-Plus、RoboCasa 基准测试中，该模型在包含复杂任务组合与动态场景扰动的设定下，平均任务成功率均达到 SOTA。其中，Libero-Plus 基 准上达到了 80.5%，较业界先进方案pi0提升近30%，展现了其在高扰动高难度具身操作任务中的领先性能。 近日，阿里巴巴集团旗下高德正式发布具身操作基座模型A ...

文丨AI研究员 卢菁博士 编辑 丨李海伦 2月10日，阿里巴巴Qwen-Image-2.0与字节跳动Seedream 5.0预览版同日登场，掀起了春节档前夕的"AI生图大战"。 这两款模型的发布引起了广泛关注，不只是因为发布时间"撞车"，更因为它们共同指向了AI生图的变化：目前，AI生图在可控生成、文字还 原、多场景适配等关键能力上，和以前相比有了明显进步。 回溯AI生图的发展轨迹，从"破圈"到"成熟"，这个赛道只用了不到四年时间。 短短几年，这个行业已从单款模型出圈演变为巨头们的亮剑竞赛，AI生图究竟经历了怎样的蜕变？曾经的"AI生图天花板"Midjourney，为何在 2026年逐渐缺位？ 本篇主要以Qwen-Image-2.0、Seedream 5.0和Nano Banana为例——前两者代表了国内头部厂商在图像生成领域的最新进展，Nano Banana则是 2025年率先打开大众市场的轻量化代表，我们聚焦这三款模型的技术路线不同点，用直接通俗的表述，讲清楚这些关键问题。 01 AI生图为什么突然"开窍"了 过去一年，AI生图完成了从"能画图"到"能干活"的质变：不再拼参数和速度，而是拼可控性、叙事能力 ...

2026年的春节，或许会被记录为中国互联网史上一个重要的分水岭。 2月12日，千问发布的"春节30亿大免单"第一波活动数据显示，过去6天，用户说了41亿次"千问帮我"， AI完成下单超1.2亿笔，涵盖了奶茶、外卖、日用百货、机酒等各类生活服务。 这组跳动的数字背后，折射出一个重要现象：中国互联网的交互逻辑正经历从"搜索获取信息"到"AI交 付服务"的代际迁徙。一个由AI驱动的、触手可及的生活新方式，已经先于预期到来。 "千问帮我"：三个月养成的新习惯 "千问帮我点两杯拿铁""千问帮我订明天去三亚的机票"——当用户自然说出这类指令时，千问App在识 别用户意图后，即可直接调用淘宝闪购、飞猪、大麦、支付宝等阿里生态服务能力，完成一个App内的 闭环交易，无需跳转页面。这种"对话即办事"的体验，正快速沉淀为用户习惯，让这个春节前夕，AI味 道十足。 随着春节人流的迁徙，新习惯正在县城中升温，让AI技术服务更多普通人。千问的数据显示，1.2亿笔 订单中有近半数AI订单来自县城。过去6天，因为AI一句话点单更便捷，156万60岁以上用户第一次体 验点外卖。 过去二十年，中国互联网的底座是"搜索+点击"。无论是PC时代的 ...

Core Insights - The "Spring Festival 3 Billion Free Order" campaign by Qianwen App has generated significant user engagement, with 4.1 billion requests for assistance and over 120 million orders completed by AI in just six days [1][5] - The campaign has notably increased consumption in county-level cities, with nearly half of the orders coming from these areas, and 1.56 million elderly users trying food delivery services for the first time [1][5] - The AI-driven shopping trend has led to a remarkable increase in retail orders, with some categories seeing growth exceeding tenfold since the campaign's launch [1][5] Group 1 - Qianwen App's campaign resulted in users ordering over 1,000 tons of eggs, more than 2,300 fitness equipment items, and over 1,500 books within six days [1] - The trend of using AI for purchasing traditional New Year goods has spurred rapid growth in instant retail orders on Taobao Flash Purchase [1][5] - Retail orders for hair dye products, gift flowers, fruit gift boxes, and emergency baby supplies have all increased by over ten times compared to the initial phase of the campaign [1][5] Group 2 - The AI technology has activated new consumption dynamics in county areas, with non-food categories like groceries and fruits seeing order growth exceeding eight times [5] - Cities such as Kunshan, Jinjiang, Nanchang, Puning, and Yiwu have emerged as hotspots for AI-driven consumption [5] - Approximately 10% of users in county areas using AI for shopping are aged 50 and above, highlighting the technology's accessibility for older demographics [5] Group 3 - The success of Qianwen App illustrates the transformative impact of AI on daily life, supported by Alibaba's comprehensive AI capabilities and ecosystem [5] - Alibaba's strategy emphasizes the integration of advanced technology with practical applications in e-commerce, instant retail, and other services [5] - The first year of AI shopping is marked by widespread consumer adoption across various age groups and regions [6]

此前的2月6日，千问"春节30亿大免单"火爆上线，连续6天霸榜苹果App Store免费榜第一名。目 前，在社交平台上，网友纷纷晒出自己用千问下单的"第一杯AI奶茶" ，人们还探索出更多千问的打开 方式：有人让千问写春联、写拜年祝福，有人让千问做PPT、写作业、翻译文件等。 2月12日上午，《国际金融报》记者获悉，当日，阿里旗下千问发布的"春节30亿大免单"第一波活 动数据显示，过去6天，用户说了41亿次"千问帮我"，AI完成下单超1.2亿笔。 淘宝闪购数据显示，在通过千问"一句话点外卖"的订单里，近半数来自县城。昆山、晋江、南昌、 普宁、义乌成为AI消费热门县。不仅如此，县城消费也正在积极拥抱"AI+即时零售"，超市百货、水果 等非餐品类订单均较活动上线初期增长超过8倍。营口盖州、丹东东港、宜春上高、运城稷山、长春榆 树等城市都有超市小店日均订单超过500单。 此外，因为AI一句话点单更便捷，数据显示，156万60岁以上用户第一次体验点外卖。全国用户通 过AI下单了1000多吨鸡蛋，还买了2300多件健身器材和1500多本图书。 根据业内人士的说法，千问爆火的背后，是阿里AI技术+生态的双轮驱动：一方面是阿 ...

上海证券指出，大模型持续升级，例如Anthropic推出Claude Opus4.6，具备"AI代理团队"功能并能 处理百万tokens上下文窗口，在金融分析等专项测试中行业领先；OpenAI发布最强编程模型GPT-5.3- Codex，在多项基准测试中表现优异。北美大厂业绩亮眼，资本开支持续加码，如Alphabet、Meta、亚 马逊和微软的营收与利润增长强劲，且均计划在2026年显著提升资本支出。国内大厂亦开启春节AI大 战，腾讯、阿里、字节和百度纷纷投入资源推广AI应用。 风险提示：提及个股仅用于行业事件分析，不构成任何个股推荐或投资建议。指数等短期涨跌仅供 参考，不代表其未来表现，亦不构成对基金业绩的承诺或保证。观点可能随市场环境变化而调整，不构 成投资建议或承诺。提及基金风险收益特征各不相同，敬请投资者仔细阅读基金法律文件，充分了解产 品要素、风险等级及收益分配原则，选择与自身风险承受能力匹配的产品，谨慎投资。 每日经济新闻 （责任编辑：张晓波 ） 创业板人工智能ETF国泰（159388）跟踪的是创业板人工智能指数（970070），单日涨跌幅限制达 20%，该指数聚焦于在创业板上市的人工智能相关企业 ...

Investment Rating - The report maintains a "Recommendation" rating for the industry, expecting the industry index to rise more than 5% over the next 3-6 months compared to the benchmark index [67]. Core Insights - The report highlights a significant transformation in the content creation paradigm due to advancements in multi-modal models, particularly with the release of new video generation models like Kuaishou's Keling 3.0 and ByteDance's Seedance 2.0, which enhance precision and controllability in video production [3][11]. - The supply-side revolution is expected to reshape the cost structure of content production, with a notable reduction in marginal costs due to improved capabilities in image and video generation [3][58]. - The report emphasizes the potential for AI video generation technology to catalyze downstream opportunities in content IP, content copyright, and AI application tools, as well as the demand for cloud services and computing power [3][58]. Summary by Sections Industry Basic Data - The industry comprises 138 listed companies with a total market value of approximately 22,562.85 billion and a circulating market value of about 20,757.76 billion [5]. Relative Index Performance - The absolute performance over 1 month, 6 months, and 12 months is 8.9%, 31.9%, and 42.7% respectively, while the relative performance is 9.6%, 16.8%, and 21.6% [6]. Model Updates - Keling 3.0 and Seedance 2.0 represent significant upgrades in video generation models, focusing on physical realism, narrative consistency, and enhanced understanding of complex text instructions [11][41]. - Keling 3.0 introduces integrated editing capabilities, allowing for controlled modifications to generated content without the need for complete regeneration [29][33]. - Seedance 2.0 enhances the precision of video production, addressing common issues such as consistency and complex action replication [41][50]. Investment Opportunities - The report identifies several investment opportunities in the content IP sector, including companies like Zhongwen Online, Yuedu Group, and Shanghai Film, as well as in content copyright and AI video production tools [58][59].

Investment Rating - The industry investment rating is "Positive" (maintained) [6] Core Insights - Domestic large models have been intensively released around the Spring Festival, initiating an AI arms race. Notable releases include DeepSeek's new model with a context processing capability of 1 million tokens, GLM-5 which ranks first globally in programming and agent testing, and ByteDance's Seedance 2.0 aimed at revolutionizing video creation [1][2] - The usability of agents is increasing, with large models transitioning from chat to collaboration. Claude Opus 4.5 can autonomously program for 5 hours, and AI coding agents are expected to double their task handling time every 4 months starting from 2024-2025, compared to a 7-month doubling period from 2019-2024 [2] - The demand for inference is expected to rise due to large-scale applications, with significant increases in token consumption for agent execution compared to dialogue scenarios. The cost of generating a 5-second 720P video is approximately 4 RMB, with Seedance costing about 2.3 RMB, indicating a substantial cost advantage over manual production [3] Summary by Sections Model Updates - MiniMax's M2.5 model is set to launch soon, currently in internal testing for the MiniMax Agent product. Other updates include GLM-5 from Zhizhu, which has achieved state-of-the-art capabilities in coding and agent functions, and DeepSeek's new model with a context window increased to 1 million tokens [7] Related Companies - Key companies mentioned include MiniMax, Zhizhu, Yunsai Zhilian, UCloud, Capital Online, Qingyun Technology, Wangsu Technology, and Nanxing Co. [4]