Core Viewpoint - The article discusses the launch of RynnBrain, the first embodied brain model with spatiotemporal memory, developed by Alibaba's Damo Academy, which significantly enhances the capabilities of embodied robots in understanding and interacting with the physical world [7][9][76]. Group 1: RynnBrain Model Features - RynnBrain consists of seven models ranging from 2B to 30B parameters, designed to understand both "time" and "space," allowing it to remember past trajectories and predict future actions [7][9]. - It outperforms leading models like Nvidia's Cosmos-reason2 and Google's Gemini Robotics ER 1.5 across 20 benchmarks, achieving 16 state-of-the-art (SOTA) results [7]. - RynnBrain-30B-A3B, the first MoE architecture in embodied models, demonstrates exceptional efficiency, requiring only 3B active parameters while surpassing the performance of a 72B model [10][11]. Group 2: Training and Data Utilization - The model was trained using over 20 million pairs of high-quality data, incorporating various multimodal training datasets to enhance its understanding of physical space [19][20]. - A unique aspect of the training involved generating 1 million pairs of "self-centered" OCR question-answer data, enabling the robot to interpret labels and numbers in its environment [21][23]. Group 3: Functional Capabilities - RynnBrain exhibits strong flexibility in input and output, capable of processing images and videos of varying resolutions and providing multiple modalities of output, such as trajectories and poses [26][28]. - It possesses spatiotemporal memory, allowing it to maintain awareness of object locations and trajectories even after interruptions, which is crucial for long-term tasks [34][40]. Group 4: System Architecture and Scalability - The model employs a "big brain-small brain" layered architecture, where RynnBrain handles long-term planning and scene understanding, while a smaller execution layer focuses on motor control [54][56]. - This architecture facilitates modular iteration and enhances the model's adaptability to various tasks, such as complex navigation and planning [57][58]. Group 5: Open Source and Industry Impact - Damo Academy has open-sourced RynnBrain along with comprehensive training codes and a new evaluation benchmark, RynnBrain-Bench, which assesses the model's understanding of video sequences and spatial positioning [60][62]. - This initiative aims to lower barriers in the industry by providing a shared infrastructure for understanding physical concepts, improving system efficiency, and fostering healthy competition among teams [66][69].

新浪科技讯 2月10日上午消息，阿里巴巴达摩院发布具身智能大脑基础模型RynnBrain，并一次性开源 了包括30B MoE在内的7个全系列模型。RynnBrain首次让机器人拥有了时空记忆和空间推理能力，智能 水平实现大幅跃升，在16项具身开源评测榜单上刷新纪录（SOTA），超越谷歌Gemini Robotics ER 1.5 等行业顶尖模型。 超越业界的72B模型效果，因此能让机器人动作更快、更丝滑。同时，达摩院还开源了全新评测基准 RynnBrain-Bench，用于评测时空细粒度具身任务，填补了行业空白。 达摩院具身智能实验室负责人赵德丽表示："RynnBrain 首次实现了大脑对物理世界的深度理解与可靠 规划，为大小脑分层架构下的通用具身智能迈出关键一步。我们期待它加速 AI 从数字世界走向真实物 理场景的落地进程。" 责任编辑：杨赐 据介绍，达摩院的RynnBrain模型创造性地引入了时空记忆和物理世界推理，这是机器人与环境互动所 需的两项基本能力。时空记忆能力可让机器人在其完整的历史记忆中定位物体、目标区域，甚至预测运 动轨迹，从而赋予机器人全局时空回溯能力。物理空间推理不同于传统的纯文本推理范 ...

Summary of Key Points from the Conference Call Industry Overview - The focus is on the AI application industry, particularly the competitive landscape among major players like Alibaba, which is seen as a leader in driving industry development [1] - The competition for AI entry points and operating systems among domestic giants is intensifying, indicating a critical phase in the industry [1] Core Insights and Arguments - Alibaba is expected to make significant advancements in model capabilities and ecosystem development by 2026, positioning itself as a key player in the AI landscape [1] - The launch of the Qianwen APP by Alibaba is anticipated to showcase the initial form of a local life agent, marking the beginning of an "administrative era" powered by AI [1] - As the agent ecosystem matures, Alibaba Cloud is projected to evolve into the Android operating system of the AI era, with aspirations to expand from domestic to global markets [1] - By the second quarter of 2026, the Qianwen APP is expected to integrate more deeply with the Taobao ecosystem, enhancing its functionality and user engagement [1] Additional Important Content - The report outlines a potential industry chain involving e-commerce, local life services, finance, advertising, healthcare, education, gaming, and judicial services, indicating a broad spectrum of applications for AI technology [1]

【阿里新一代模型Qwen3.5曝光】财联社2月9日电，在全球最大AI开源社区HuggingFace的开源项目页面中，最新出 现Qwen3.5并入Transformers的新PR（提交代码合并申请）。 这意味着阿里千问新一代基座模型Qwen3.5或发布在即。 相关信息透露，千问3.5采用了全新的混合注意力机制，并且极有可能是原生可实现视觉理解的VLM类模型。 【阿里新一代模型Qwen3.5曝光】财联社2月9日电，在全球最大AI开源社区HuggingFace的开源项目页面中，最新出 现Qwen3.5并入Transformers的新PR（提交代码合并申请）。 这意味着阿里千问新一代基座模型Qwen3.5或发布在即。 相关信息透露，千问3.5采用了全新的混合注意力机制，并且极有可能是原生可实现视觉理解的VLM类模型。 有开发者进一步挖掘出，Qwe 有开发者进一步挖掘出，Qwen3.5或将开源至少2B的密集模型和35B-A3B的MoE模型。 这一最新动态印证了此前有消息称Qwen3.5将在春节期间开源。 ...

Summary of Key Points from the Conference Call Industry Overview - **AI Industry**: Major companies like Tencent and Alibaba are launching significant AI applications for the upcoming Spring Festival, indicating a competitive landscape in AI technology and applications [1] - **Commercial Space and Solar Power**: The commercial space industry is approaching a critical phase in 2026, with successful launches of reusable spacecraft in China. Tesla is reportedly expanding its solar battery manufacturing footprint across multiple states in the U.S. [2] - **Robotics**: Various companies are showcasing their robotics technologies during the Spring Festival, with a notable emphasis on the importance of dexterous robotic hands as highlighted by Elon Musk [3] - **AI Animation and Computing Power**: ByteDance has quietly launched its AI video generation model Seedance 2.0, which utilizes breakthrough technologies. The Ministry of Industry and Information Technology in China is promoting a national computing power interconnectivity system [4] - **Stablecoins**: The Hong Kong Monetary Authority is preparing to issue the first batch of stablecoin licenses by March, with a focus on maintaining stability in the issuance process [5] Core Insights and Arguments - **AI Applications**: The competition among major tech firms in AI applications is intensifying, with significant investments and innovations being made to capture market share during high-traffic events like the Spring Festival [1] - **Commercial Space Developments**: The advancements in reusable spacecraft technology are expected to revolutionize the commercial space sector, potentially leading to increased investment and growth opportunities [2] - **Robotics Innovation**: The emphasis on dexterous robotic hands suggests a shift towards more sophisticated robotic capabilities, which could enhance automation across various industries [3] - **AI and Computing Power**: The launch of Seedance 2.0 signifies a major advancement in AI capabilities, which could disrupt the video content creation industry and enhance user engagement [4] - **Stablecoin Regulation**: The cautious approach to stablecoin licensing in Hong Kong reflects a broader trend towards regulatory oversight in the cryptocurrency space, aiming to ensure market stability [5] Additional Important Content - **Collaborations in AI**: Companies like Huya and Kuaishou are partnering with AI cloud service providers for major events, indicating a trend towards collaborative efforts in enhancing AI capabilities [1] - **Investment Opportunities**: The developments in commercial space and solar power sectors present potential investment opportunities, particularly in companies involved in these technologies [2] - **Market Trends in Robotics**: The participation of various robotics companies in high-profile events suggests a growing market interest and potential for innovation in this field [3] - **Government Initiatives**: The push for a national computing power interconnectivity system indicates government support for technological advancements, which could benefit related industries [4] - **Stablecoin Market Dynamics**: The limited issuance of stablecoin licenses may create a competitive environment among a few players, impacting market dynamics and investment strategies [5]

2月初港股科技陷入波动，恒生科技指数周跌6.5%，招商证券策略分析师张夏指出：只是剧烈的流动性 冲击。往后看，做多港股科技的赔率与胜率均较高。 逢低布局港股科技相关ETF：恒生科技指数ETF(513180.SH)、恒生互联网ETF(513330.SH)、港股通科技 ETF基金(159101.SZ)。其中，恒生互联网ETF(513330.SH)聚焦港股互联网巨头(阿里巴巴/百度集团/腾讯 控股/网易)；港股通科技ETF基金(159101.SZ)在恒生科技基础上，龙头权重更集中(阿里巴巴/腾讯控 股)，且补齐港股特色创新药板块(百济神州/药明生物/信达生物)。 上述ETF均在内地证券交易所上市，并支持T+0日内灵活回转交易，为A股投资者提供低门槛、免开跨 境户和免换汇的便利。 从赔率来看：如果用恒生科技指数/A股双创指数来衡量科技板块AH溢价，港股科技相对于A股科技的 折价接近历史最高水平。上一次触及低点还是2022年3月，当前对互联网公司的监管与经济发展环境明 显好于2022年和2023年，在AI大发展与科技兴国的时代背景下，显然港股科技已经被显著低估。 从胜率上看：利好的因素在不断聚集。从流动性看，海外流动性 ...

【申万宏源通信】多模态需求+春节应用推广，看好算力底座/IDC成长(20260210早） #字节多模态Seedance2.0加码算力需求。 Seedance2.0在分镜运镜、音画同步、多镜头叙事能力等方面进步显著；一旦多模态渗透率提升，对tokens和算力底 座需求量将指数级增长。 年底年初字节、阿里等招投标进展，初步推动IDC估值修复。 但中长期真正打开 【申万宏源通信】多模态需求+春节应用推广，看好算力底座/IDC成长(20260210早） #字节多模态Seedance2.0加码算力需求。 Seedance2.0在分镜运镜、音画同步、多镜头叙事能力等方面进步显著；一旦多模态渗透率提升，对tokens和算力底 座需求量将指数级增长。 腾讯链： 年底年初字节、阿里等招投标进展，初步推动IDC估值修复。 但中长期真正打开智算中心需求的，是云厂应用/agent/多模态的商业化进展。 伴随CSP春节期间持续营销投流打开泛化AI入口，IDC作为算力底座真实需求将显著提升，预计行业持续高景气。 #从选股的角度，字节、阿里、腾讯等后续有望轮动表现。 看好绑定云厂的核心IDC。 字节链（优先推荐）： #东阳光（字节核心、远 ...

编辑 | 邢昀 元老们回归掌控阿里，定下了AI和电商两大战略方向，目前这两个方向已经形成了一个清晰的交汇 点：奶茶。 文 | 豹变 张经纬 前有2025年夏天，淘宝闪购在外卖大战中，靠几块钱甚至免单的奶茶/咖啡，从美团、京东手中抢外卖 份额。后有千问APP在春节撒币30亿，邀请全国人民免费喝奶茶，为AI造势。 根据阿里2月6日的战报，红包发出后，9小时内千问订单超1000万，其中绝大多数是奶茶/咖啡。看起 来，阿里试图复制闪购的经验，在AI用户争夺战中后来居上。 不管是近场电商的争夺战，还是培育AI用户，阿里的抓手都离不开奶茶，这场以奶茶为武器的豪赌前 路如何？真能帮千问在用户心中扎根吗？ 01 "我是马云，给我点一杯奶茶" 从用户的反应来看，发券的效果是明显的：2月6日这天一大早，各路人马就开始在各个群里转发口令， 之后点奶茶的人太多，千问和运力就都扛不住了。 在朋友圈和小红书，很多人都在吐槽千问崩了。同样崩了的还有阿里的股价，开盘下跌4%，经过一天 的动荡后以3%的跌幅收盘。而港股茶饮股大多应声而涨，古茗涨幅3.58%，股价创上市以来新高，茶百 道收盘涨超6%，沪上阿姨、蜜雪冰城也有不同程度的上涨。 阿里 ...

过去一周多时间，许多朋友跟我辩论AI的入口问题。尤其当百度文心，腾讯元宝先后开启春节红包大战之后，市场和 舆论越发倾向于"入口本质上是市场费用投放的较量"：对比过往历次春节红包大战，很容易得出"谁能在短期内有足够 的资金去"烧"，谁就拿到胜利的筹码"的结论。 但阿里千问却别出心裁，没有选择红包大战，而是率先开启了一场AI生活和购物的一次全民科普。2月6日，千问App 正式上线"春节30亿大免单"活动，且上线仅9小时，App内的下单量突破1000万单。之后不断有朋友来跟我辩论：这到 底是不是"预算"的胜利？ 那么我们该如何看待AI入口大战以及阿里千问这次现象级"爆表"呢？本文核心观点： 其一，在美国全栈式AI的谷歌获得极大利好，与微软的下行形成强烈对比，国内全栈AI的代表——阿里正在走谷歌道 路； 其二，2026年初千问App融合阿里零售生态，阿里股价开始异于谷歌，市场非常关注阿里AI应用的新探索； 对比美国行业发展之后，我们再把目光放在我国大模型厂商，截止目前在AI的产业布局上，阿里是最接近谷歌的（全 世界唯二的全栈AI公司）。 其三，千问App现象级刷屏只是多年前阿里的"因"种成的"果"。 阿里和谷歌：全世 ...

2026年开年，教育再次成为了巨头们觊觎的赛道。 1月中旬，字节旗下豆包APP，将AI学习工具"豆包爱学"推上核心入口；几乎同时，阿里旗下千问APP新增"一键搜试卷"功能，全国各名校试卷可直接下 载，直指传统教育应用腹地。 教育这个曾因"双减"而降温的市场，随着大模型的介入，又重新热闹起来。QuestMobile数据显示，截至2025年第三季度，国内AI教育应用月活已突破1.2 亿，同比增长340%。 一场围绕"AI+教育"的争夺战已然打响。对阵双方旗帜鲜明：一方是坐拥C端流量的字节、阿里等互联网大厂；另一方则是有教研积淀的作业帮、猿辅 导、好未来等传统教育玩家。 大厂入局的逻辑很清晰：教育需求明确，市场广阔，而大模型技术恰好能复用。但大厂想从教育市场里分一杯羹，仅靠堆砌功能和流量思维，恐怕还远远 不够。 1 大厂靠流量，教培靠教研 当前AI教育赛道主要汇聚了三股势力：一是以字节、阿里等大厂为代表的技术派，凭借大模型底座与流量入场，追求快速迭代和场景验证；二是以猿辅 导、作业帮、好未来为代表的教培派，核心战略是用AI技术强化内容和教育服务，实现存量用户的转化；三是中小创业公司，它们避开了与巨头的正面 竞争，在 ...