证券研究报告 Manus即将加入Meta， AI Agent应用推广未来将有望加速 计算机行业 强于大市（维持） 平安证券研究所计算机团队 分析师： 闫磊 S1060517070006（证券投资咨询）YANLEI511@pingan.com.cn 2026年1月4日 请务必阅读正文后免责条款 黄韦涵 S1060523070003（证券投资咨询）HUANGWEIHAN235@pingan.com.cn 刘云坤 S1060525120004（证券投资咨询）LIUYUNKUN518@pingan.com.cn Manus即将加入Meta，AI Agent应用推广未来将有望加速 事件描述：北京时间12月30日，Manus发布公告称，Manus即将加入Meta。 点评：Manus称，这是对Manus在通用AI Agent领域里工作的认可。根据Manus公告信息，自发布以来，Manus 专注于构建通用型 AI Agent，帮助用户高效完成研究、自动化和复杂任务。面对全球越来越多用户的使用需求，团队持续迭代产品，努力使 Manus 在实际使用中更实用、更可靠。根据2025年12月初统计的数据，上线至今，Manus 已处理超 ...

行业热点 雷军：未来五年研发至少投入 2000 亿元，加大大模型投入 整理 | 傅宇琪、褚杏娟 雷军：未来五年研发至少投入 2000 亿元，加大大模型投入；微信回应安装包 10 多年膨胀数百倍；Anthropic 豪掷 210 亿美元购 谷歌 100 万块 TPU；比亚迪首次超越特斯拉，成全球最大电动汽车销售商；技术元老离场！腾讯 AI Lab 副主任俞栋离职；传快 手副总裁、基础大模型及推荐大模型负责人周国睿即将离职；新论文暗示 DeepSeek V4 已完成训练；Manus 武汉团队基本搬 离，核心业务人员迁往新加坡；"全球大模型第一股"来了！智谱发行市值达 511 亿港元；50 亿美元联姻！NVIDIA 正式收购 Intel 股份…… 1 月 3 日晚，小米集团董事长雷军在新年首场直播中透露，小米汽车 2026 年全年交付目标为 55 万辆，2025 年交付量超 41 万辆，超过原先计划的 30 万辆。而在今日早间发布的微博中，雷军称，"希望今年也能超额完成（目标）。" 雷军介绍小米规划聚焦三点：一是未来五年至少投入 2000 亿元坚持技术研发；二是加大对大模型的投入；三是坚持为人车家 全生态体验打造极 ...

责任编辑：王进和 专题：中国财富管理50人论坛2025年会 1月4日金融一线消息，中国财富管理50人论坛2025年会近日在京召开，本届年会的主题是"迈向'十五 五'建设金融强国"。在前沿对话环节，腾讯集团副总裁韩开创表示，大模型正从技术工具演变为金融 的"核心生产力引擎"，驱动服务从"人找服务"走向"服务懂人"，并深刻重塑投研、风控与产品形态。金 融机构实现这一变革的关键，在于构建"通用"与"垂类"能力的深度协同：以通用模型为"通才"处理意图 与规划，而将机构独有的业务逻辑与数据注入垂类"专家"，从而避免"千模一面"，构建难以复制的差异 化优势。善用AI的机构，将定义未来金融。 责任编辑：王进和 专题：中国财富管理50人论坛2025年会 1月4日金融一线消息，中国财富管理50人论坛2025年会近日在京召开，本届年会的主题是"迈向'十五 五'建设金融强国"。在前沿对话环节，腾讯集团副总裁韩开创表示，大模型正从技术工具演变为金融 的"核心生产力引擎"，驱动服务从"人找服务"走向"服务懂人"，并深刻重塑投研、风控与产品形态。金 融机构实现这一变革的关键，在于构建"通用"与"垂类"能力的深度协同：以通用模型为"通才" ...

继壁仞科技之后，又来一个新ai融资，理论上是2个，智谱和minimax。 智谱： 1.8日上市！发行价：116.2港元/股，全球发售3741.95万股H股，募资净额约41.734亿港元，IPO估值约518亿港元（约478亿人民币）。 新公司，过去几轮估值： 天使轮 2019年 未披露 君联资本等投资 A轮 2021年9月 未披露 融资过亿元，达晨财智等参投 B轮 2022年9月 25.8亿元 融资数亿元，君联、启明领投 C轮 2024年6月 210亿元 融资4亿美元，Prosperity7领投 C+轮 2024年8月 未披露 融资超10亿元，君联领投，腾讯等跟投 D轮 2024年12月 未披露 财务： 公司 收入来自通过 MaaS 平台提供大模型服务。该平台向用户开放矩阵式模型能力以及智能体工具套件，并支持本地化部署及云端部署等多种方式灵活 部署。 2022-2024年整体营业收入分别为0.57亿元、1.25亿元、3.12亿元，2022年至2024年营收复合年增长率达133.28%。 截至2025年上半年公司营收达1.91亿元，较2024年同期大幅增长325.0%， 其中本地化部署为营收核心支柱，营收大幅增长 ...

Investment Rating - The industry is rated as "Overweight," indicating an expected increase in value exceeding 10% compared to the market benchmark index over the next 6-12 months [6]. Core Insights - The large model has entered a post-training and COT expansion phase, with the capabilities of the base model in 2025 likely remaining unchanged, as the GPT-5 series may still utilize the GPT-4o base model. The focus for 2025 will be on enhancing post-training and reasoning capabilities [1]. - Google's Gemini 3 model has achieved significant advancements in cross-modal dialogue, understanding, and content generation, but still faces challenges with logical coherence in complex scenarios and controllability of generated content, highlighting key areas for future technological breakthroughs [1]. - Domestic AI chip manufacturers have reached H-series performance levels, with advancements in interconnect speeds and software ecosystem capabilities. Notably, Alibaba's latest PPU chip has surpassed NVIDIA's A800 in key performance metrics, and Huawei's CloudMatrix 384 super node aims to optimize computing efficiency [2]. - The capabilities of robotic bodies have improved significantly, while the cognitive abilities of their "brains" lag behind, limiting their application to structured scenarios. The VLA model architecture has been criticized for its limitations in real-time reasoning in complex physical environments [3]. - The business models for AI applications are still under exploration, with domestic companies facing challenges in monetization despite high revenue growth rates, while international firms struggle with high computing costs and low profit margins [3].

Core Insights - DeepSeek has introduced a new neural network architecture optimization called mHC (Manifold-Constrained Hyper-Connections), which is expected to significantly impact the AI industry, including large models and chips [1][5][9] Group 1: mHC Architecture - The mHC architecture builds on the Hyper-Connections (HC) framework released by the Byte Bean team in November 2024, aiming to replace the nearly decade-old ResNet architecture [5] - mHC introduces a Manifold-Constrained approach using the Sinkhorn-Knopp algorithm to stabilize signal propagation during training, addressing issues of signal explosion and instability in large model training [5][6] - In training demonstrations with 27 billion parameters, mHC maintained a signal amplification of only 1.6 times, while HC experienced a catastrophic failure with a 3000 times amplification [6][8] Group 2: Performance and Efficiency - mHC shows a significant reduction in training loss and improved performance on challenging tasks, with over 2% enhancement in reasoning and reading comprehension benchmarks compared to traditional architectures [6][8] - The additional training time overhead for mHC, even with a fourfold expansion of residual channels, is only 6.7%, indicating a focus on cost-effectiveness and efficiency [8] Group 3: Industry Impact and Reactions - The release of mHC has sparked high discussion levels among researchers and industry professionals, with expectations of a paradigm shift in large model architectures by 2026 [9][10] - Competitors are already responding, with new architectures like Deep Delta Learning emerging shortly after mHC's announcement, indicating a potential chain reaction in AI architecture development [9][10] - Analysts predict that DeepSeek may make significant announcements around the Lunar New Year, potentially unveiling the long-awaited R2 model or a faster universal model V4 [10] Group 4: Compatibility and Market Dynamics - mHC's architecture is primarily designed for NVIDIA's supernode links, raising concerns about compatibility with domestic chips, which may require enhanced adaptation efforts [11] - As U.S. AI chip manufacturers gradually exit the Chinese market due to geopolitical factors, domestic chipmakers are accelerating their development and ecosystem building to adapt to DeepSeek's models [12]

同时，他提到，和千问的团队很多成员都比较熟悉。在季逸超看来，千问是做的非常的扎实，而且它是 真正第一个上非常宽松开源的模型，对此他非常尊敬。另外，千问让他看到了大厂里一群年轻人能干出 来事情的表现，非常难得。 他解释到，自己很早就和Deepseek有接触。"我之前的一次创业是在做知识图谱构建，且同时我们自己 从零搭建整个搜索引擎的Infra，有很大的预训练数据集，所以当时很早就跟Deepseek的朋友认识了。但 后来我也没有把数据单独卖给他们，我直接把数据集开源了，也算是给大家留下一点遗产。" 此外，他还提到，Kimi一方面也是真格基金投资的，因为自己也在真格，所以当时交流比较多。另一 方面，季逸超认为，Kimi这个公司是比较有品位的，这个很重要。 新浪声明：所有会议实录均为现场速记整理，未经演讲者审阅，新浪网登载此文出于传递更多信息之目 的，并不意味着赞同其观点或证实其描述。 近日，在访谈中，谈及大模型公司，Manus联合创始人兼首席科学家季逸超表示，自己比较喜欢的有三 家——千问、Deepseek和Kimi。 同时，他提到，和千问的团队很多成员都比较熟悉。在季逸超看来，千问是做的非常的扎实，而且它是 真正 ...

数字人民币重大调整,利好金融 IT 和三方支付 [Table_Industry] 综合金融 ——金融科技行业双周报第二十期 | [姓名table_Authors] | 电话 | 邮箱 | 登记编号 | | --- | --- | --- | --- | | 刘欣琦(分析师) | 021-38676647 | liuxinqi@gtht.com | S0880515050001 | | 孙坤(分析师) | 021-38038260 | sunkun@gtht.com | S0880523030001 | 本报告导读： 数字人民币迎利好，允许数币付息和部分准备金制度落地、跨境支付政策落地，利 好金融 IT 和三方支付；部分金融科技行业持续严格监管。 投资要点： 综合金融《数币余额开始付息，供需双向促规模 增长》2026.01.03 综合金融《数币体系升级，规模有望加速增长》 2025.12.29 综合金融《大模型第一股将上市,百融云发布 RaaS 战略》2025.12.22 综合金融《投融资再平衡持续，股权融资逐步回 暖》2025.12.22 股 票 研 究 研 究 报 告 请务必阅读正文之后的免责条款部分 行 ...

小米汽车2026年目标定了。 今日（1月4日）早间，小米创始人、董事长兼CEO雷军在社交平台发文称，"小米汽车，去年交付目标 30万辆，后来提升到35万辆，最终全年交付超过了41万辆。昨晚直播，我公布了2026年交付目标：55万 辆。希望今年也能超额完成。感谢车主朋友们支持！" 谈到2026年的计划，雷军表示，小米将聚焦三个方面。一是坚持技术立业，过去五年小米在技术研发上 差不多投了1050亿元，未来五年至少会投2000亿元。二是加大对大模型的投入，用AI全面赋能。三是 坚持为用户打造"人车家全生态"的体验，把产品做到极致。 就在前一日晚间，雷军举办了新年第一场直播，该直播以工程师现场拆解全新小米YU7为主要形式展 开，并回应了部分网络谣言。 雷军在直播中称，欢迎更多的博主来拆解小米的汽车。他同时表示，人不可能违背良心，去说一些假的 内容。"特别希望大家能说一些公道话，不要为了流量去故意夸大，故意找茬。"雷军说。 直播中，雷军坦言，过去一段时间扑面而来的质疑，让他们有点懵。其中不少是恶意的抹黑，包 括"7000法务""不让农民卖小米""绿化带战神"等，甚至他自己发健身的微博也会被骂。网络尬黑也给他 造成了很大 ...

Core Viewpoint - The article discusses the strategic decisions made by the company "月之暗面" (Kimi) regarding its funding and market approach, emphasizing its preference for private funding over public listing at this stage, while also outlining its ambitious plans for technological advancements and product development [5][6][8]. Funding and Market Strategy - The company recently completed a $500 million (approximately 3.5 billion RMB) Series C funding round, significantly exceeding its initial target, with a post-money valuation of $4.3 billion [7]. - Notable investors in this round included IDG, Alibaba, Tencent, and Meituan's co-founder Wang Huiwen, highlighting strong market confidence in the company [7]. - The CEO, Yang Zhilin, stated that the company is not in a hurry to go public, believing it can raise more capital through private markets than through IPOs [6][8]. Technological Advancements - The funding will be used to aggressively expand GPU resources and accelerate the training and development of the K3 model, aiming to enhance its capabilities significantly [10]. - The company plans to achieve a tenfold increase in the effective FLOPs of the K3 model and introduce unique features that differentiate it from competitors [10]. - Recent product launches include new intelligent agent functionalities and the Kimi-K2 model, which boasts 1 trillion parameters, positioning the company among a select few capable of developing such large models [12][13]. Commercialization and User Growth - Kimi has implemented a tiered subscription model for its services, resulting in a significant increase in paid user growth, with a monthly average increase of over 170% from September to November [15]. - The release of K2 Thinking has led to a fourfold increase in overseas API revenue during the same period [15]. Company Culture and Leadership - The company fosters a unique and efficient culture characterized by direct communication among top talent, allowing for high productivity with smaller teams [16][19]. - CEO Yang Zhilin, a highly accomplished individual with a strong academic background, is considered a key asset to the company [19]. - The company has shifted its focus from aggressive marketing to enhancing its foundational algorithms and model capabilities, maintaining a long-term strategic vision despite short-term user engagement fluctuations [19].