去年，横空出世的 DeepSeek 在春节假期突袭了整个科技界。 数以亿计返乡的中国人，在漫长的假期里捧着手机，见证了 DeepSeek R1 如何在大洋彼岸的硅谷引发震动。 【导读】这个春节，中国 AI 迎来「决战时刻」。据《The Information》援引内部消息人士透露：字节或将祭出全模态三件套；阿里除了或将发布强大的 全新旗舰模型 Qwen 3.5 外，也会让千问打通支付与电商，挑战豆包；DeepSeek V4 或将携最强代码能力突袭。这不仅是技术竞赛，更是对 14 亿用户生活 入口与未来互联网秩序的终极争夺。 对于中国的互联网巨头而言，即将到来的春节注定无法平静。 那场突如其来的爆红，不仅让一家公司成为了民族骄傲，更重写了中国 AI 产业的竞争节奏。 如今，又是一年春节将至。 当数亿人准备再次陷入假期那种「有大把时间盯着屏幕」的状态时，中国科技圈的巨头们已经决定不再给对手留下任何独占舞台的机会。 一场关于 14 亿人未来如何生活、工作与思考的「群雄逐鹿」，正在字节跳动、阿里巴巴和 DeepSeek 之间悄然拉开帷幕。 字节跳动：三款多模态旗舰模型 在字节跳动的办公区，气氛比往年更加紧绷。 这家以 ...

Group 1 - The market showed signs of recovery on January 30, with the AI application sector's decline narrowing, as the CSI Software Service Index fell by 2.5% [1] - The software ETF E Fund (562930) has attracted significant market attention, with net inflows exceeding 150 million yuan over six consecutive trading days [1] - DeepSeek is set to release its next-generation flagship AI model, DeepSeek V4, in February, which is expected to have a substantial impact on the current AI competitive landscape [1] Group 2 - Bohai Securities believes that the iteration and upgrade of domestic large models like DeepSeek V4 will accelerate technological innovation and application, maintaining a positive outlook on the growth potential of the AI application sector driven by both technological advancements and demand release [1] - The E Fund software ETF tracks the CSI Software Service Index, which consists of 30 stocks involved in software development and services, covering various AI application scenarios such as AI + office, AI + finance, and AI + education [1] - The top ten weighted stocks in the index include iFlytek, Kingsoft Office, Tonghuashun, Guiding Compass, and Hengsheng Electronics, collectively accounting for over 60% of the index, providing investors with a convenient tool for investing in the AI application sector [1]

2026年1月30日A股回调整理，截至午间收盘，中证软件服务指数下跌3.06%。成分股方面，深信服领 跌，中科星图、广联达、卫宁健康、朗新科技跟跌。 近期，DeepSeek即将发布V4大模型，其编程能力被预期将对标并超越Claude与GPT系列闭源模型。中原 证券指出，V4在编码能力、长上下文理解及训练稳定性方面实现质变式跃迁，且依托mHC新架构与 Engram条件记忆模块，在降低训练成本的同时缓解GPU内存瓶颈，有望显著改善国产AI芯片HBM受限 局面，并加速国产算力生态建设。 软件ETF(159852)跟踪中证软件服务指数，是掘金计算机软件行业的便利工具。 场外投资者还可以通过软件ETF联接基金（012620）布局AI软件投资机遇。 华创证券指出，智能体（Agent）应用爆发正推动CPU需求结构升级，域外CPU在多线程调度、低延迟 响应与能效比方面成为性能决定性环节。IDC预计活跃Agent数量将从2025年的2860万飙升至2030年的 22.16亿，年复合增速达139%，叠加AIDC扩容与先进制程产能倾斜，服务器CPU已进入紧平衡格局，国 产龙头有望凭借定制化能力切入增量市场。 数据显示，截至2025 ...

Core Viewpoint - Recent updates from multiple domestic large model manufacturers indicate a shift from merely competing on parameters and dialogue performance to a deeper focus on engineering and system-level capabilities, aiming to transition large models from "research achievements" to "industrial products" [1] Group 1: Model Updates - Alibaba released the Qwen3-Max-Thinking flagship reasoning model, while DeepSeek and Kimi updated their models with DeepSeek-OCR 2 and Kimi K2.5 respectively [1] - MiniMax launched the Music2.5 music generation model, addressing two major AI music technology challenges, which significantly boosted stock prices in the Hong Kong market, with MiniMax's stock rising over 20% and Zhiyu's stock increasing over 10% [1] Group 2: Challenges in Engineering Phase - The first challenge is balancing cost and efficiency, as high-parameter models incur substantial training and inference costs, making it financially burdensome for most companies to adopt top models for full-scale business operations [2] - The second challenge involves meeting industrial-grade requirements for stability and interpretability, as current models still exhibit issues like "hallucinations" and output variability, which could pose significant risks in critical applications such as financial risk control and medical diagnosis [2] - The third challenge is the integration with existing systems, which requires complex API connections, data format conversions, workflow restructuring, and adaptation of security frameworks, yet many models remain at the "chat demonstration" level without deep integration capabilities [2] Group 3: Path to Overcoming Challenges - Breakthroughs in each challenge are technically demanding, necessitating a shift from "pursuing extreme parameters" to "optimizing unit computational efficiency" to ensure affordability and usability for enterprises [3] - Companies are increasingly seeking stable problem-solving capabilities rather than just technical specifications, prompting a shift from merely providing models to offering comprehensive services and solutions [3] - Implementing techniques like prompt engineering and retrieval-augmented generation can help build safeguards for key application scenarios, effectively controlling hallucinations and enhancing result reliability and interpretability [3]

漫剧 " VA 1月29日，360正式宣布国内首个工业级AI漫剧智能体生产平台——"纳米漫剧流水线"启动公开测试。该平台专为漫剧工作室、视频制作公司及专业创作者量 身打造，旨在破解行业"效率与品质不可兼得"的核心痛点，实现"高效量产、品质可控、风格独特"的精品AI漫剧生产新体验，开启AI漫剧工业化生产的新时 代。 线 纳米漫剧流水线 ♀ 案例广场 高效量产，操作便捷：制作速度达到主流工具的3倍以上，单集生产时间可压缩至30分钟至1小时；素材生成成功率突破90%，远超行业平均水平，制作精品 漫剧从此不需要浪费时间抽卡。通过"流水线推进+智能画布调整"双模式交互，实现精品内容的规模化高效产出。 目前，纳米漫剧流水线已与保利影业、华视娱乐、奇想文化、友和文化、兴艺凯晨、九紫源AI等多家头部影视及漫剧公司达成合作，共同探索基于工业化 流程的漫剧生产新模式，以系统性提升内容质量与产能。值得一提的是，已有部分合作项目完成制作，将于近期陆续上线。其中九紫源AI是爆款金奖AI拟 真人剧《机甲局》的制作方，据悉双方将合作推出《机甲局》第二季。 纳米漫剧流水线不仅是一款生产力工具，更是360推动构建AI漫剧行业标准的重要载体，体 ...

Core Insights - Two prominent domestic AI startups, DeepSeek and Kimi, have released significant open-source updates to their models, DeepSeek-OCR 2 and K2.5, respectively, marking a pivotal moment in AI development [1][4] - DeepSeek-OCR 2 focuses on enhancing the model's ability to "read" information through a new visual encoding mechanism, aiming to improve efficiency and reliability in processing complex documents [1][10] - Kimi K2.5 aims to evolve AI from merely answering questions to executing complex tasks, emphasizing long memory, multi-modal understanding, and task execution capabilities [4][12] Group 1: DeepSeek-OCR 2 - DeepSeek-OCR 2 introduces a new approach to document processing, allowing the model to learn human-like visual logic and compress lengthy text inputs into higher-density "visual semantics" [1][10] - The model shifts from a mechanical text processing method to understanding document structure, enabling it to identify titles, tables, and related information more effectively [8][10] - This upgrade addresses long-standing issues in AI document handling, such as high costs and inefficiencies associated with traditional text input methods [10][11] Group 2: Kimi K2.5 - Kimi K2.5 emphasizes the transition from a question-answering model to a more capable digital assistant, capable of handling complex tasks and multi-modal inputs [4][12] - The model's long memory feature allows it to retain context over extended interactions, reducing the need for repeated explanations [12][17] - Kimi K2.5's focus on task execution and intelligent agent capabilities positions it as a more versatile tool for real-world applications, moving beyond simple advisory roles [12][22] Group 3: Industry Trends - The recent upgrades in AI models reflect a broader industry shift towards practical applications, prioritizing usability and integration into real-world workflows over mere parameter scaling [15][16] - Key areas of focus include enhancing memory retention, improving visual comprehension, and redefining AI's role from advisor to executor [17][22] - The emphasis on engineering and deployment capabilities highlights the industry's commitment to making AI tools more accessible and effective in business environments [22][23]

1. 中国人工智能领域的下一个关键突破，或将来自本土对标英伟达的芯片企业的技术迭代 2. 近数月来，摩尔线程、沐曦、壁仞、燧原这四家中国芯片初创企业已完成上市或提交上市申请 3. 专家表示，华为旗下芯片的性能已接近英伟达同类产品 DeepSeek颠覆美国市场一年后，投资者正做好准备，迎接中国人工智能领域的下一个关键突破：一批 本土对标英伟达的人工智能芯片企业将协同发力、加速规模化发展。 就在过去两个月，被称作中国人工智能芯片 "四小龙" 的四家初创企业 —— 摩尔线程、沐曦、壁仞、燧 原，均已完成上市或正式提交上市申请。 这四家企业的发展，均助力中国实现更宏大的产业战略：完善本土人工智能硬件产业链，降低对美国技 术的依赖。 核心要点： 与此同时，华为、寒武纪等老牌芯片企业也在快速扩大产能。 在去年 9 月举办的年度全联接大会上，华为发布了三年内赶超英伟达的发展规划。 这家科技巨头拥有实现这一目标的技术积淀与产业经验：其打造的全球通信业务版图已成规模，在国内 市场从苹果手中抢占了大量智能手机份额，还在短短数年间跻身中国云计算市场第一梯队。 但高端芯片仍是人工智能产业链中最难攻克的环节之一。 人工智能计算初创企业 ...

Core Insights - Google DeepMind has introduced a significant new capability called Agentic Vision for Gemini 3 Flash, transforming how large language models understand the world from passive guessing to active investigation [1][3][5]. Technology Overview - Agentic Vision allows models to actively manipulate images based on user requests by employing a "Think-Act-Observe" loop, enhancing the model's ability to analyze and interact with visual data [3][11]. - This capability has resulted in a performance improvement of 5% to 10% across various visual benchmark tests for Gemini 3 Flash [6]. Practical Applications - The technology enables developers to unlock new behaviors through code execution in the API, demonstrated in applications like PlanCheckSolver.com, which improved accuracy by 5% through iterative checks of high-resolution inputs [10]. - Agentic Vision facilitates image annotation, allowing the model to interact with the environment by drawing and labeling directly on images, ensuring pixel-level accuracy in its responses [13]. - The model can also perform visual mathematics and plotting, generating visual representations of data while avoiding common pitfalls of standard large language models [15][16]. Future Prospects - Google indicates that Agentic Vision is just the beginning, with plans to enhance implicit actions like image rotation and visual mathematics in future updates, as well as exploring additional tools for the Gemini model [20]. Competitive Landscape - The release of Agentic Vision coincides with DeepSeek's launch of DeepSeek-OCR2, suggesting a competitive response in the field of visual AI, where both companies are redefining machine vision capabilities [21][22]. - The competition centers around who can better define machine vision, with DeepSeek focusing on perception and Google emphasizing interactive capabilities through code execution [23].

当开年以来AI界都在讨论智能体和AI员工的话题时，DeepSeek在1月27日悄无声息地发布了一项看似"无聊"的更新——DeepSeek-OCR2。 作为2025年10月20日DeepSeek-OCR1发布后时隔三个月的最新更新，尽管看似并不如传说中的V4让人期待，但是DeepSeek-OCR2的公布可能直接敲响了 OCR(文档识别)这个千亿级市场的丧钟。 过去十年，OCR一直是一门隐秘而暴利的生意。从Adobe的PDF编辑器，到扫描全能王的会员费，再到亚马逊AWS Textract昂贵的API调用，无数公司靠 着"教机器认字"赚取了丰厚的利润。 这种模式的死穴在于"不懂逻辑"。遇到报纸的跨栏排版，它会把两篇无关的文章拼在一起；面对扭曲的发票，就找不到对齐线；对于密集的小字财报，只 能看到模糊化的文本。 而DeepSeek-OCR2引入了"视觉因果流"的概念。在DeepEncoder-V2中，研究团队用一种类语言模型结构替代了原先基于CLIP的视觉编码模块，并在编码器 内部引入可学习的"因果流查询token"。 其编码器同时包含双向注意力与因果注意力两种处理模式，原始视觉信息通过双向注意力进行全局感知，而新增 ...

与此同时，国内人工智能初创公司月之暗面Kimi正式发布了新一代开源模型 Kimi K2.5。据介绍，Kimi K2.5基于原生多模态架构设计，支持视觉与文本输 入，将视觉理解与推理、编程、Agent等能力全部集成到一个模型当中。 今日（1.28），港股科技午后大涨，100%纯度的港股通科技ETF汇添富（520980）涨近1%，成交额超3.6亿元！资金已连续3天大幅涌入，累计"吸金"超4.8 亿元！ 消息面上，1月27日，DeepSeek团队发布《DeepSeek-OCR 2: Visual Causal Flow》论文，并开源DeepSeek-OCR 2模型，采用创新的DeepEncoder V2方法， 让AI能够根据图像的含义动态重排图像的各个部分，更接近人类的视觉编码逻辑。 另外，阿里巴巴于1月26日晚间正式发布千问旗舰推理模型Qwen3-Max-Thinking，在多项关键性能基准测试中，千问表现超过了GPT-5.2、Claude Opus 4.5 和Gemini 3 Pro等顶尖模型，刷新全球纪录，进一步拓展了AI系统的推理性能边界。 海外方面，云厂商迎涨价潮。当地时间1月27日，谷歌云正式官宣涨价。 ...