据介绍，作为首部聚焦通用大模型的国家标准，该系列标准填补了技术评价体系空白，明确性能、安全 与服务能力要求，配套评测能力已获中国合格评定国家认可委员会（CNAS）认可。 截至目前，标准工具完成千余项评测任务，调用大模型超95万次，精准识别幻觉控制、内容安全等共性 问题，助力近30家厂商技术迭代，推动形成"研发—评测—应用—升级"闭环。 新华社北京12月26日电（记者 赵文君）记者12月26日从市场监管总局获悉，人工智能大模型系列国家 标准正式实施，标志着我国大模型产业进入"科学权威、统一规范"新阶段。 ...

中经记者 曲忠芳 北京报道 智谱方面称，亏损主要是由于对研发作出重大投资。在报告期内，智谱研发投入逐年增加，2022年至 2024年分别是8440万元、5.29亿元、21.95亿元，占收入的百分比分别是147%、425%、703%。2025年上 半年研发支出为15.95亿元，占总收入的百分比高达835%，而2024年上半年这一百分比数字为1913%。 三年半的时间里累计研发投入超过44亿元。 12月19日，港交所官网显示，北京智谱华章科技股份有限公司（以下简称"智谱"）已通过聆讯并公布了 IPO（首次公开募股）招股书等资料，不仅成为业界所称的"大模型六小龙"中率先步入IPO流程的企 业，而且也抢在OpenAI、Anthropic等海外大模型明星企业前面，有望成为"全球大模型第一股"。随着 488页招股书的披露，公众终于得以一窥智谱大模型生意的"成色"。 需要指出的是，就在12月21日一早，"大模型六小龙"中另一家MiniMax也在港交所披露了聆讯资料集。 《中国经营报》记者注意到，不同于此前多数企业在递表时会先公示一个"申请版本"的招股书，智谱与 近日通过聆讯的壁仞科技、天数智芯等，均是直接披露聆讯后资料集， ...

Core Viewpoint - The articles emphasize the importance of the artificial intelligence (AI) industry and the need for proactive engagement from all stakeholders to seize historical opportunities in this rapidly evolving field [1][2][3]. Group 1: AI Industry Development - The AI industry is characterized by the interconnection of big data, computing power, and large models, with large models being pivotal for driving the entire industry forward [2]. - A provincial-level opinion on accelerating the high-quality development of the large model industry is being finalized, reflecting a commitment to support industry pioneers [2][3]. - The government is shifting from broad funding approaches to targeted support for the core needs of the large model industry, introducing innovative policies like model vouchers and corpus vouchers [3]. Group 2: Achievements and Recognition - At a recent awards ceremony, 30 AI large model companies were recognized for their contributions, with leading enterprises demonstrating mature technology systems that support various sectors such as chemicals and government [4]. - Successful applications of AI have been reported, including a 30% increase in production efficiency through smart optimization and a rise in product quality from 95% to 99.8% using intelligent inspection models [5][6]. - The AI industry is witnessing significant advancements, with both major and small enterprises contributing to innovative solutions that address industry challenges and enhance productivity [6]. Group 3: Collaborative Efforts - The collaboration between government and enterprises is crucial for fostering innovation and ensuring the successful implementation of AI technologies [6]. - The articles highlight the belief that joint efforts will lead to the realization of the potential within AI, ultimately driving high-quality development in the industry [6][7].

齐鲁晚报·齐鲁壹点记者 周雨婷 时培磊 该实验室立足山东丰富的文旅资源禀赋，整合集团在大模型研发、数字复刻、智能交互等方面的技术优 势，致力于推动数智文旅领域关键技术的突破与成果高效转化，具体聚焦四大核心研究方向：文旅行业 依托齐鲁文化大模型，集团联合尼山世界儒学中心、孔子研究院等单位共建上线孔子数智大模型；联合 山东沂蒙精神研究会、山东省委党校等单位，建设上线沂蒙精神数智大模型；联合山东省作家协会，建 设上线山东文学大模型。正在推进齐鲁文化遗产大模型、中医药文化大模型等项目建设。此外，搭建山 东文化数据标注平台，与人民网落地建设全国首个主流文化语料库。 | 序号 | 依托单位 | 实验室名称 | | --- | --- | --- | | 1 | 山东省艺术研究院 | 艺术资源数智转化场景山东省文化和旅游重点 | | | | 实验室 | | 2 | 山东省文物考古研究院 | 科技考古与文物保护山东省文化和旅游重点 实验室 | | | 山东省图书馆(山东省古籍 | 古籍保护与修复山东省文化和旅游重点实验室 | | 3 | 保护中心) | | | র্ব | 山东旅游职业学院 | 人工智能赋能旅游行业山东省文化 ...

高歌猛进与巨额亏损并存 当前冲刺港股的大模型公司，呈现出几个鲜明的共同特征： 1. 营收高速增长，但亏损规模惊人 根据披露的财务数据，智谱AI与MiniMax均处于"增收不增利"的典型状态。智谱AI在2024年实现营收 3.12亿元，但净亏损高达29.58亿元；2025年上半年营收1.91亿元，净亏损23.58亿元，三年半累计亏损超 过62亿元。其亏损核心在于巨额的研发与算力投入，仅2024年研发开支就达21.95亿元，占营收的 703%。Min 近日，中国人工智能领域迎来标志性事件：两家头部大模型公司——北京智谱华章科技股份 有限公司（智谱AI）与稀宇科技（MiniMax）——几乎同时通过港交所上市聆讯，一场关 于"大模型第一股"的资本竞速赛进入最后冲刺阶段。 这不仅是一场简单的IPO竞赛，更标志着中国AI产业从早期的技术狂热与资本追捧，正式迈入接受公开 市场审视、考验真实价值创造能力的新阶段。 现状扫描： ...

Core Insights - The article discusses the evolving identity of Zhipu AI as it transitions from a traditional ToB service provider to a significant player in the global AI model competition, particularly in the context of its IPO and the rapid advancements in its model capabilities [2][3][16]. Group 1: Company Overview - Zhipu AI's revenue from localized deployment accounts for 85%, which has led to a narrative of being an AI outsourcing company focused on delivery rather than API services [3]. - The company has shifted its commercialization strategy from "model delivery" to "model invocation," aiming for a balanced revenue stream between localized deployment and model-as-a-service (MaaS) [2][8]. - CEO Zhang Peng indicated that the future revenue split between localized deployment and MaaS will be 50-50, reflecting a strategic pivot in their business model [2]. Group 2: Market Position and Competitiveness - Zhipu AI has been compared to OpenAI, not just in product form but in its foundational technology, positioning it within a global competitive framework for large models [2][16]. - The company has released multiple iterations of its GLM model (GLM-4.5, 4.6, and 4.7) in rapid succession, establishing itself as a key player in the developer ecosystem [3][5]. - Zhipu AI's models have gained traction among global developers, with significant adoption and integration into mainstream AI platforms, indicating its transition from a domestic alternative to a global competitor [8][10]. Group 3: Financial Performance and Growth - The company reported a significant increase in paid developer subscriptions, surpassing 150,000 within two months of launching its GLM Coding plan, leading to an annual recurring revenue exceeding 100 million [8]. - In November, Zhipu AI achieved a daily token invocation volume of 4.2 trillion, countering the narrative of being merely a delivery-focused company [9]. - The overall gross margin has shown improvement as model invocation scales up, indicating a positive trend in financial performance despite initial losses [13]. Group 4: Industry Context and Future Outlook - The AI market is expected to grow significantly, with estimates suggesting that the Chinese B-end AI market could reach 100 billion RMB by 2030, highlighting the potential for Zhipu AI's growth [13]. - The article emphasizes that the valuation of Zhipu AI should not be based solely on past financial data but rather on its potential as a foundational model provider in the evolving AI landscape [16][17]. - Zhipu AI's strategic focus on building trust with enterprise clients through localized deployments and its independent model development is seen as a competitive advantage in the global market [15][20].

来源：新浪证券 出品：新浪财经上市公司研究院 作者：IPO再融资组/郑权 近日， 国内大模型公司智谱华章（下称"智谱"）、稀宇科技MINIMAX（下称"MINIMAX"）先后披露 了港股招股书，争夺"大模型第一股"。 根据招股书及wind数据统计，智谱的营收略高，MINIMAX募资额多但亏损额较大（注：主要以2024年 财务数据对比，MINIMAX的美元数据按照历史汇率以人民币数据统计，下同） 。两家公司的研发人员 占总员工数量比相差不多，但智谱的研发人员数量是MINIMAX的两倍多，MINIMAX的研发费用率、 国内发明专利数量、著作权数量也比智谱少。 招股书显示，MINIMAX 的语言模型Speech-02-HD曾排名全球第一，OpenAI、Google都排在其后面， 这与大部分投资者的普遍认知有差距。试问，MINIMAX 招股书选择的评价机构、排名数据是否有参考 价值还值得考证。 MINIMAX 融资金额多但亏损金额高 资料显示，智谱与MINIMAX同属"中国AI六小龙"的明星企业。两家公司的定位和商业模式有明显差 异：智谱致力于开发先进的通用大模型，构建了全面的MaaS平台，通过"开源基座+商业定制" ...

炒股就看金麒麟分析师研报，权威，专业，及时，全面，助您挖掘潜力主题机会！ 出品：新浪财经上市公司研究院 作者：IPO再融资组/郑权 近日， 国内大模型公司智谱华章（下称"智谱"）、稀宇科技MINIMAX（下称"MINIMAX"）先后披露了港股招股书，争 夺"大模型第一股"。 根据招股书及wind数据统计，智谱的营收略高，MINIMAX募资额多但亏损额较大（注：主要以2024年财务数据对比， MINIMAX的美元数据按照历史汇率以人民币数据统计，下同） 。 另一方面，MINIMAX的现金储备要比智谱多，这对大模型企业而言至为关键，因为研发需要继续"烧钱"，更好的模型需要 大量的训练。同时，智谱的国际化收入也不及MINIMAX，后者境外收入占比约七成，智谱的境外收入占比较低，收入基本 上都来自中国内地。 智谱最值得关注的问题应该是大客户变动问题。2023年度前四大客户B、C、D、E，2022年度前两大客户F、G，报告期内交 易很短时间就难觅踪影。2025年上半年，公司机构客户减少2400多家，减少比例超43%。 智谱现金储备不及MINIMAX 国际化收入逊色 资料显示，智谱与MINIMAX同属"中国AI六小龙"的 ...

Core Insights - 2025 is designated as the "International Year of Quantum Science and Technology" by UNESCO, marking a pivotal moment in China's "14th Five-Year Plan" and a shift in computational paradigms towards quantum computing and advanced materials [1][22] Quantum Computing Breakthroughs - The DeepSeek-R1 model, released on January 20, 2025, utilizes reinforcement learning to enhance reasoning capabilities and achieves performance comparable to OpenAI's models at a low computational cost [3][24] - The Nobel Prize in Physics awarded in October recognizes the foundational role of superconducting quantum circuits in modern computing, highlighting the work of John Clarke, Michel H. Devoret, and John M. Martinis [5][26] - The Helios quantum computer, launched by Quantinuum, features 98 physical qubits with a fidelity of 99.9975% for single-qubit operations, making it the most precise commercial quantum computer to date [9][30] Advanced Materials and Technologies - The BISC chip, developed by a collaboration between Columbia University and Stanford, integrates 65,536 electrodes into a flexible CMOS chip, enabling unprecedented wireless data transmission rates of up to 100 Mbps [8][29] - A new type of hollow-core optical fiber, reported by the Lumenisity research team, demonstrates a loss of only 0.091 dB/km at a wavelength of 1550 nm, promising advancements in long-distance communication [11][32] - The development of "atomic manufacturing" techniques has pushed metal materials to the Ångström scale, opening pathways for new quantum and electronic devices [13][34] Environmental and Energy Innovations - SSAB's Zero™ steel has become the first product to meet the near-zero emissions threshold set by the International Energy Agency, utilizing hydrogen reduction technology to eliminate carbon emissions [17][38] - The EAST nuclear fusion experiment achieved a world record of 1 billion degrees Celsius for 1066 seconds, demonstrating the feasibility of future fusion reactors [18][39] - Research on the Bennu asteroid samples revealed the presence of essential life-building components, suggesting that the building blocks of life may have been distributed across early Earth and other planets [21][42]

Core Viewpoint - The Nasdaq 100 index has risen over 20% this year, with semiconductors and network infrastructure leading, while telecom and payment software lag behind. Looking ahead to 2026, the index is expected to show steady returns, particularly in the first half, due to AI spending concerns creating low expectation opportunities [1] Group 1: Key Issues Influencing Tech Stocks in 2026 - The debate on AI is shifting focus from computing power to physical AI (robotics, autonomous driving) and how regulation and return on invested capital (ROIC) will evolve [1] - Valuation recovery in the software industry will be influenced by the end of seat pricing models, the rise of intelligent agents, and competition in large language model (LLM) commercialization [1] - Apple's strategic positioning will be questioned: whether it will act as a defensive growth stock or an AI narrative vehicle, and if foldable smartphones can catalyze growth [1] - The supercycle in commodities, particularly in DRAM, memory chips, and precious metals like gold and copper, will raise questions about who can absorb cost pressures [1] - The paradox of generative AI efficiency, driven by layoffs enhancing productivity, may exacerbate non-farm employment data pressures [1] - Meta and similar companies will be seen as having the most investment value in the debate over profit margins and competition [1] - The potential for cyclical industries such as housing, commercial real estate, analog chips, or automotive sectors to hit bottom will be analyzed [1] - Conditions for hardware stocks to lead will depend on discussions around gross margins and visibility of spending potentially suppressing semiconductor market trends [1] - The development path of LLMs in the race for artificial general intelligence (AGI) will focus on whether US and Chinese models will move towards productization or primitive intelligence competition [1] - Potential blind spots include whether the return of agency businesses or SaaS stocks will become a consensus in 2026 [1]