2026年1月8日，港股市场迎来一场资本狂欢。智谱华章（股票代码：02513.HK）以210港元高开，较发行价暴涨80.7%，盘中市值突破900亿港元，成为全球 AI大模型领域市值最高的非美国上市公司。这场资本盛宴的背后，是中国AI大模型产业对OpenAI全球霸主地位的首次正面挑战。当智谱的AutoGLM模型在 港股敲钟声中同步完成第150万次手机操作任务时，一个关键问题浮出水面：中国AI大模型，是否已具备改写全球技术格局的实力？ 资本市场已用脚投票。2025年港股AI板块融资额同比增长177%，智谱、MiniMAX等头部企业募资总额超200亿港元。这种资本聚集效应正在形成正向循 环：资金推动技术迭代，技术突破吸引更多资本，最终构建起独立于西方技术体系的AI生态。 智谱华章的上市表现堪称现象级。招股首日孖展认购超58倍，冻结资金超290亿港元，最终上市首日涨幅达110%，市值直逼千亿港元。这种热度与2024年 OpenAI估值突破万亿美元时华尔街的疯狂如出一辙，但不同之处在于：智谱的崛起路径完全独立于西方技术体系。 其核心武器GLM-4.7模型在代码生成领域已实现对GPT-5.2的超越。在HumanEval基 ...

编者按：AI产业的竞争不是比谁跑得更快，而是比谁跑得更稳、更远。2025年，人工智能正以势如破 竹之势重塑未来。从DeepSeek开源大模型的火热"出圈"，到多场人形机器人赛事"极限破圈"，再到"千 问速度"引爆科技圈，AI正以前所未有的速度穿透虚拟与现实、串联技术与产业。站在AI技术革命的拐 点，我们正目睹一场从技术奇迹到实用价值的深刻范式转移，资本市场对人工智能的追逐已从概念炒作 转向价值重估。站在2025年末，新华财经从上游芯片和算力、中游大模型研发、下游端侧应用等角度纵 观这一年的非凡答卷，也以理性视角审视这场浪潮中的真实与泡影，探寻中国人工智能产业升级的关键 路径。 新华财经北京12月29日电（可达） 2025年无论是在资本市场上还是生活中的方方面面，人工智能都占 据着最重要的"风口"。其中，热度最高的无疑就是"大模型"，从"深度求索（DeepSeek）"当选年度国内 词，到"大模型"入选年度十大科普热词，都无一例外地证明了这一点。 这一年，作为大模型发展的关键之年，其发展已然超脱出技术上的范畴，开始成为人们身边的智能助 理，办公室、生产线上的"生产力"工具。智能变革的大幕正徐徐拉开。 而在岁末更替 ...

两家公司虽同属大模型赛道，但成长路径与产品方向存在明显分野。 智谱源自清华大学计算机系的技术成果转化，由该系知识工程实验室的团队创立。公司专注于AGI的研发，其基于自回归填空的全国产预训练架构GLM系 列模型已适配40余款国产芯片，被业内视为通用性最高的模型之一。 两种路径 作为"大模型六小虎"中首家启动IPO流程的公司，智谱最早在2025年4月即在北京证监局办理辅导备案。而MiniMax的上市进程同样备受关注，据信其目标 是在2026年1月登陆资本市场。 12月17日，智谱通过港交所上市聆讯，并于12月19日正式披露招股书。紧随其后，MiniMax也完成了相同程序，其招股书于12月21日首次刊发。 当两只AI独角兽几乎同时通过港交所上市聆讯，一场关于"全球大模型第一股"的资本盛宴，正式拉开帷幕。 智谱构建了全面的产品矩阵，覆盖从边缘小模型到百亿参数旗舰大模型的不同规模。其产品功能涵盖对话、通用智能体、代码生成、图像理解、文生图/ 视频及语音交互等多个场景。 在冲刺IPO之前，两家公司已在资本市场上获得了充分的弹药储备。 智谱自2019年成立以来，已完成多轮融资。投资者中既有高瓴资本、启明创投、君联资本等知名 ...

图片来源：AI生成 上个月，埃隆·马斯克（Elon Musk）在知名视频播客节目《乔・罗根体验》（The Joe Rogan Experience）中说，未来5到6年内，人们所熟知的传统智能手机 及独立的手机应用程序将会消失。 未来的设备将不再是独立的智能终端，而会成为一个"AI推理的边缘节点"，它主要承担显示、音频播放和网络通信的功能，复杂的思考和生成任务将由云端 AI与本地AI协同完成。 马斯克话音未落，大洋彼岸关于AI手机的生态之战就已经悄然打响。 12月初，豆包联合中兴推出努比亚M153工程机，被称为"豆包AI手机"，原价3499元，首批3万台已售罄，在二手平台该手机全新未拆封版本报价3999到4999 元不等，最高报价炒至3.6万元，甚至还衍生出日租金50到79元不等的短期租赁业务。 事实上，AI手机并不是什么新概念。早在2017年前后，三星、苹果、华为、小米等头部厂商就已提出"AI手机"的概念，并推出过相关机型。 但早期手机厂商的AI功能多聚焦在单点辅助。比如，苹果曾提出的AI手机概念，包含让Siri制定旅游计划、订机票等，但国行机型相关功能迟迟未落地，且 首版AI Siri表现未达预期。 其他厂 ...

Core Insights - The AI industry is experiencing an accelerated competition for terminal entry points, particularly in mobile devices, with major players like Zhiyu and ByteDance making significant advancements in AI capabilities [1][2][3] Group 1: AI Model Developments - Zhiyu has open-sourced its AutoGLM model, aiming to make AI capabilities accessible for the entire industry, thereby lowering entry barriers [2][5] - ByteDance's Doubao team has introduced a mobile assistant that integrates deeply with the operating system, showcasing capabilities such as executing tasks based on user commands [2][3] Group 2: Industry Challenges - Both companies acknowledge that the performance of AI agents in real-world scenarios is still far from perfect, with significant challenges remaining in model intelligence and task execution [3][5] - Concerns regarding user privacy and data security are paramount, especially as AI systems gain access to sensitive applications like payment software [3][5] Group 3: Strategic Implications - The competition for AI super entry points is not limited to mobile phones; it is expanding to wearable devices and native applications, indicating a strategic battle for the next generation of traffic entry [6][7] - Companies like Xiaomi and Alibaba are exploring new hardware forms, such as AI glasses, to redefine user interaction with digital environments [6][7] Group 4: Future Directions - The evolution of applications into super AI applications is underway, with Alibaba focusing on both AI for businesses and consumers, potentially transforming how users interact with services [8] - The shift towards AI agents as super entry points raises questions about the future value of traditional app advertising and the willingness of major apps to share data with AI systems [8]

Core Insights - The concept of "Agent" has transitioned from being a topic of debate to a critical focus in the AI industry, as evidenced by its prominence at WAIC 2025, where over 800 companies showcased more than 3000 exhibits, doubling previous years' participation [1][2] Trend Summaries Trend 1: Agents as a Necessity - The term "Agent" has become ubiquitous across various exhibitors, indicating a widespread recognition of its importance in AI applications [2] - Siemens showcased its Industrial Copilot system, which integrates AI to enhance industrial processes, demonstrating the practical application of Agents in real-time operations [4] Trend 2: Evolution of AI Capabilities - AI is evolving from a mere chat tool to a more creative and productive tool, with companies like MiniMax highlighting the shift towards Agents that can perform complex tasks autonomously [5] - The AutoGLM model from Zhiyu AI exemplifies this trend by autonomously executing various tasks, indicating a move towards more interactive and capable AI systems [5] Trend 3: Multi-Agent Collaboration - The shift from single-agent systems to multi-agent collaboration is seen as a key to tackling complex tasks, with companies demonstrating how multiple Agents can work together to enhance efficiency [7] - The transition from "tool thinking" to "collaborative partner thinking" reflects a deeper integration of AI capabilities into business processes [7] Trend 4: Results Over Services - The focus has shifted from showcasing features to delivering tangible results, with companies prioritizing practical solutions that meet user needs [9][11] - MiniMax's Agent demonstrates the ability to execute tasks efficiently, highlighting the importance of outcome-oriented AI solutions [9] Trend 5: Rise of Consumer Products - The explosion of consumer-oriented AI products at WAIC 2025 signifies a new phase in AI development, where Agents are recognized as essential software products in the digital landscape [14] - WPS Lingxi, a standout product, showcases the ability to facilitate document creation through natural language processing, emphasizing user-friendly AI applications [14] Trend 6: Infrastructure Development for Agents - The foundational infrastructure for Agents is being strengthened, with companies like Alibaba Cloud introducing solutions like "Wuying AgentBay" to streamline AI development [16] - PPIO's launch of an Agentic AI infrastructure service platform aims to lower technical barriers for developers, facilitating broader adoption of AI technologies [17]

Core Insights - The WAIC showcased significant advancements in AI and robotics, with over 350,000 attendees participating in the event [1] - Major tech companies shifted focus from basic large models to multi-modal applications and AI agents, indicating a competitive landscape [1][3] - The emergence of AI agents as a primary focus for companies, with various solutions being demonstrated across different sectors [6][7] Group 1: Event Overview - WAIC attracted approximately 350,000 attendees, highlighting the growing interest in AI and robotics [1] - The event featured over 800 exhibitors showcasing advancements in AI infrastructure, robotics, and multi-modal applications [1][2] - The shift from traditional humanoid robots to more interactive and functional robots was evident, with live demonstrations of various tasks [10][11] Group 2: Company Highlights - Alibaba's booth was the largest, featuring the Quark AI glasses and multiple open-source large models, emphasizing their commitment to AI agents [3][6] - Ant Group presented various AI solutions, including the financial reasoning model Agentar-Fin-R1, showcasing their focus on industry-specific applications [6][7] - The "Six Little Tigers" of large models showed a divergence in performance, with some companies like Baichuan Intelligence and Zero One falling behind [7][8] Group 3: Technological Developments - The AI agents market has surpassed $5 billion, with a growth rate of 40%, indicating a strong demand for practical applications [4] - Companies are increasingly focusing on integrating AI models with real-world business needs, as seen in the development of solutions for document proofreading and financial services [5][6] - The introduction of advanced components like six-dimensional force sensors is enhancing the capabilities of humanoid robots, allowing them to perform complex tasks autonomously [12][14] Group 4: Market Trends - The trend is shifting from "technology showcase" to "scene rehearsal," with a focus on practical applications of AI technology [14] - The competition is intensifying as companies strive to effectively integrate technology into products, moving beyond mere demonstrations [14] - The rapid growth in the humanoid robot market is creating both challenges and opportunities for component manufacturers, necessitating faster development cycles and higher standards [13][14]

Core Viewpoint - The WAIC 2025 highlights the evolution of AI from a tool to a partner, emphasizing the need for a new coexistence relationship as AI becomes more human-like and capable [2][45]. Group 1: Trends Observed at WAIC 2025 - The event showcased six major trends in AI development, including the emergence of collaborative robots, AI as a personal efficiency partner, and the evolution of AI agents from chat tools to task executors [2][4][30]. - The exhibition featured over 800 companies and 3000 exhibits, indicating a clear path for AI development in the post-large model era [2]. Group 2: Embodied Intelligence - Embodied intelligence is transitioning from a display phase to practical applications, with robots capable of executing tasks in real-world scenarios, such as product retrieval and sorting [4][5]. - Companies like Beijing Galaxy General and Shanghai Zhiyuan are leading in this space, showcasing robots that can interact and perform complex actions autonomously [4][5]. Group 3: AI as a Personal Efficiency Partner - Major companies like Huawei, Alibaba, and Baidu are demonstrating how AI can enhance personal capabilities, with Huawei's Ascend 384 super node significantly improving data transmission efficiency [9][10][16]. - Alibaba's Quark AI glasses represent a new form of human-computer interaction, integrating navigation and payment functionalities [12][16]. - Baidu's "Miao Da" platform allows non-technical users to create applications easily, showcasing AI's potential to democratize technology [13][15]. Group 4: Evolution of AI Agents - AI agents are evolving from simple chat tools to sophisticated task executors, with platforms like Baidu's NOVA enabling digital assistants to perform complex tasks in real-time [18][20]. - The MiniMax Agent demonstrates capabilities in understanding requirements and executing tasks across various applications, indicating a shift towards more functional AI systems [20][24]. Group 5: Startup Landscape - The "AI Six Dragons" have shown a diversification in strategies, with companies like Jiyue Xingchen and MiniMax presenting clear product strategies and technical capabilities [23][24]. - New entrants like Flowith are gaining attention for their innovative approaches to AI interaction, focusing on non-linear thinking and collaborative problem-solving [41][44]. Group 6: Human-Centric AI - AI is increasingly being designed to connect with human emotions and enhance work efficiency, as seen in products from companies like SenseTime and Kingsoft [30][32][33]. - The shift from AI as a mere tool to a more human-like partner is evident, with products designed to engage emotionally and assist in complex tasks [36][38]. Group 7: Emerging AI Stars - New startups like Flowith are emerging as significant players in the AI interaction space, focusing on innovative user experiences and rapid product development [41][44].

Core Insights - The WAIC 2025 showcased over 800 companies and 3000 exhibits, highlighting a clearer path for AI development in the post-large model era [2] - Six key trends were identified, indicating a shift from basic AI capabilities to high-frequency applications in service, office, content, and interaction [2] Group 1: Trends in AI Development - Trend 1: Embodied intelligence is moving from a display phase to practical applications with collaborative robots capable of executing tasks in real-world scenarios [3][4] - Trend 2: AI is becoming an "invisible partner" for personal efficiency, with major companies like Huawei, Alibaba, and Baidu showcasing AI's role in enhancing individual capabilities [6][18] - Trend 3: AI agents are evolving from chat tools to task execution, with platforms like Baidu's NOVA enabling advanced digital human interactions [19][26] Group 2: Company Innovations - Huawei's Ascend 384 super node was highlighted for its significant improvements in resource scheduling and computational performance, enhancing AI model training efficiency [8] - Alibaba introduced the Quark AI glasses, marking a shift in AI applications from software to multi-form daily devices [10] - Baidu's "Miao Da" platform allows non-technical users to create complex applications, demonstrating AI's potential to streamline development processes [14][18] Group 3: Emerging AI Startups - New startups like Flowith are gaining attention for their innovative approaches to AI interaction, focusing on non-linear thinking and collaborative creativity [56][58] - The "AI Six Dragons" are diversifying, with companies like Step and MiniMax showcasing distinct product strategies and technological capabilities [27][29] Group 4: Human-AI Interaction - AI is transitioning from a tool to a partner, with products like SenseTime's emotional companion robots and Kingsoft's WPS Lingxi assistant enhancing user engagement and productivity [42][44][55] - The trend indicates a deeper integration of AI into personal and professional environments, emphasizing emotional connection and collaborative efficiency [55][62]

Core Insights - The company, Zhiyuan (Beijing Zhiyuan Huazhang Technology Co., Ltd.), officially launched the intelligent agent "AutoGLM Rumination" at the Zhongguancun Forum on March 31, showcasing its deep research capabilities and practical operations, marking a transition to "thinking while doing" in AI agents [1] - CEO Zhang Peng highlighted the presence of a Scaling Law in agents, indicating that by expanding inference compute during training, agents demonstrate enhanced performance [1] - The technology evolution path of AutoGLM includes the GLM-4 base model, GLM-Z1 inference model, GLM-Z1-Rumination model, and the AutoGLM model, with core models and technologies set to be open-sourced on April 14 to promote industry ecosystem development [1] Model Development - Based on recent technological advancements, the company retrained a base model called GLM-4-Air-0414 with 32 billion parameters, incorporating more code and reasoning data during the pre-training phase and optimizing for agent capabilities during the alignment phase, significantly enhancing its abilities in tool invocation and online search tasks [2]