顶级AI实验室如OpenAI、Google、xAI、Microsoft，都抢着把自家模型送来「试镜」。 一场AI界的《创造101》火了！LMArena让你盲投选出最强AI，三年从校园项目逆袭，刚刚融1.5亿美元，估值飙到17亿美元。众包投票挑战专家权威， 争议四起，却已成行业标杆。你的票，就能决定下一个AI顶流！ 一场AI界的《创造101》火了！ ChatGPT、Claude、Gemini、Grok等一众AI「练习生」整齐站好，紧张等待公演。 这不是选秀节目，而是真实发生在lmarena.ai上的AI大乱斗。 这个曾经的校园开源小项目，最近刚融到1.5亿美元，估值17亿美元。 现在，AI强弱不再是大公司自己说了算，决定权掌握在全球网民手上。 这场「AI创造营」到底怎么玩？谁会成为下一个出道顶流？让我们一起揭秘。 LMArena的「选秀起源」，从校园项目到硅谷舞台 一切要从2023年说起。当时，加州大学伯克利分校Sky Computing Lab里，一群研究生和教授搞了个开源小项目，叫Chatbot Arena。 创始人包括伯克利计算机科学教授Ion Stoica（Databricks联合创始人）、研究生Ana ...

杨秀玲介绍，2025年预计北京市人工智能核心产业规模预计达到4500亿元，企业集聚超2500家，这两项关键 指标均占全国半数左右。在头部企业培育方面，北京已拥有人工智能领域上市企业近60家、独角兽企业约40 家，并诞生了国内市值最高的人工智能独角兽、单笔融资额最大的具身智能企业与科学智能企业，以及首个 上市的国产人工智能芯片企业和大模型企业。在应用层面，现象级产品持续涌现。2025年三季度，字节跳动 旗下"豆包"月活跃用户数突破1.72亿，成为国内月活规模最高的原生AI应用，手机助手功能更带动行业生态 变革。 杨秀玲透露，下一步，北京将深入实施"九大行动"，加快建设全球人工智能创新高地，力争两年左右北京人 工智能核心产业规模突破万亿元。 人工智能已成为推动现代社会发展的核心引擎，它通过深度学习和大数据处理能力，正在重塑医疗健康、科 学研究、工业生产、教育文化和日常生活等各个领域。AI不仅能高效完成重复性工作，解放人类创造力， 还能在疾病诊断、药物研发、气候预测等复杂问题上提供突破性解决方案，其重要性体现在既是经济增长的 新动能，也是应对人类重大挑战的关键工具，更是决定未来国际竞争力的战略性技术。 我国高度重视 ...

Core Insights - The "Beijing Artificial Intelligence Innovation High Ground Construction Action Plan" was officially released, marking a new phase in Beijing's journey to become the "AI Capital" [1][8] - By 2025, the core AI industry in Beijing is projected to reach a scale of 450 billion yuan, with over 2,500 companies, accounting for approximately half of the national total [1][8] Industry Development - Beijing has made significant progress in implementing the "AI+" action plan, leading to the initial emergence of industrial cluster effects and a complete AI industry chain covering foundational, technical, and application layers [3][10] - As of mid-2025, the core AI industry in Beijing reached 215.22 billion yuan, reflecting a year-on-year growth of 25.3% [3][10] - The city is home to nearly 60 listed AI companies and about 40 unicorns, showcasing a robust industrial ecosystem [3][10] Innovation Ecosystem - Beijing ranks second globally in AI innovation cities, with notable advantages in talent and research, including 2 national laboratories and 14 universities with AI programs [5][12] - The city is fostering deep integration of AI technology into urban development, focusing on both industrial empowerment and public welfare through innovative mechanisms [5][12] Application Areas - AI technologies are significantly empowering key industries such as manufacturing, finance, automotive marketing, energy, and agriculture [7][14] - Notable projects include the launch of the "Qingrui" model by Jingneng Group for precise monitoring and risk prediction, and the "Yuanzheng Intelligent" system by BAIC Group covering all dimensions from technology to user experience [7][14] Action Plans - The action plan outlines nine key initiatives to create a comprehensive development framework, including technology innovation, infrastructure enhancement, data resource development, and application promotion [8][14] - Future goals include a substantial increase in original capabilities in AI theory and core technologies, with plans to attract over 200 billion yuan in investment funds and to see the emergence of more than 10 new listed companies and 20 unicorns in the AI sector [8][15]

智通财经APP获悉，近日，范式智能（06682）宣布，旗下「信创模盒」ModelHub XC已完成对智谱 （02513）共计26款主流模型的全面适配与深度优化。此次适配覆盖多模态对话、文本生成、文生视频等 丰富AI任务类型，进一步强化了信创模盒在国产算力环境下即开即用的模型生态能力。 未来，范式智能将继续携手智谱等优秀模型方，以及海光、沐曦等算力伙伴，深化软硬协同优化，持续 拓展模型生态边界。公司旨在通过「信创模盒」这一枢纽平台，携手更多生态伙伴，共同推动AI技术在 各行各业实现安全、高效、规模化的应用，为构建成熟、开放的国产AI基础设施生态持续贡献力量。 智谱是一家国内大模型领域的重要公司，即将于1月8日在香港联交所主板挂牌上市。其业务专注于多模 态对话、文本与视频生成等技术的商业化，为企业与开发者提供AI解决方案与平台服务。 范式智能是一家2014年成立的上市公司，是中国AI+企业服务领域的领导者，2025年实现集团化升级后， 公司主营业务涵盖面向企业服务的B端业务第四范式，以及面向消费电子市场的C端业务Phancy。 范式智能与智谱的技术协同，是国产软硬件生态融合的又一实践。此次与智谱模型的深度融合，不 ...

Group 1 - The current market offers a unique opportunity for investors to gain exposure to private, venture-style tech investments without the need for traditional brokerage accounts or stock trading [2] - The Fundrise Innovation Fund serves as a bridge between venture capital and mutual funds, targeting small, long-term investors [3] - This fund is a closed-end, non-diversified tender-offer fund that allocates capital to private technology companies, focusing on artificial intelligence, software, and modern data infrastructure [4] Group 2 - The minimum investment for the Innovation Fund starts at around $10, significantly lower than traditional venture funds that typically require six-figure commitments [5] - The fund operates with limited liquidity, as shares do not trade on an exchange and investors can only exit during periodic tender offers [5][6] - The portfolio includes notable private companies such as OpenAI, Anthropic, Databricks, and Canva, which are well-capitalized and central to current AI and software infrastructure trends [7] Group 3 - Ownership of shares in the Innovation Fund does not equate to owning stock in any single company, as performance is based on the overall portfolio rather than individual company success [8]

Group 1 - The core viewpoint of the article highlights the challenges and opportunities faced by large model companies, particularly focusing on their transition towards a more stable business model centered around API services for B2B clients [2][3][10] - The article discusses the significant interest in IPOs for large model companies, with notable subscription rates for companies like Zhipu and MiniMax, indicating a strong market appetite [1] - It emphasizes the competitive landscape, where companies like Anthropic are leading the enterprise-level LLM API market, with a projected 32% market share by 2025, and the need for domestic companies to adapt to this trend [2][15] Group 2 - Zhipu's business model is shifting from localized deployment to a focus on API services, aiming to increase the revenue share from API business to 50% [4][9] - The financial performance of Zhipu shows a concerning trend, with net losses increasing significantly from 1.44 billion in 2022 to 29.58 billion in 2024, and a projected loss of 23.58 billion in the first half of 2025 [19][21] - The article outlines the challenges faced by Zhipu in achieving profitability, with a negative gross margin for its cloud deployment business and high R&D costs primarily driven by computing power expenses [5][14][21] Group 3 - The competitive environment in the domestic market is described as a "red ocean," with price wars becoming a significant factor as companies strive to capture market share [22][26] - Zhipu's strategy includes integrating its G2B and B2B operations to streamline resources and improve efficiency, reflecting a broader trend among large model companies to focus on core capabilities [27][29] - The article concludes that the ability to convert R&D investments into stable cash flow will be a critical test for all large model companies as they navigate the transition to public markets [29]

Photographer: Gabby Jones/Bloomberg Two of China’s largest generative AI startups are set for back-to-back debuts in Hong Kong this week, testing investor appetite for an industry that blossomed in the post-ChatGPT era. Yet Zhipu and MiniMax — among the country’s earliest challengers to sector leaders including OpenAI — took very different routes to public markets. Zhipu and Minimax begin trading Thursday and Friday, respectively, in a litmus test of whether investors believe China’s fledgling AI indust ...

Core Insights - DeepSeek has introduced a new architecture called mHC aimed at addressing stability issues in large-scale model training while maintaining performance improvements [1][11]. Group 1: Problem Identification - Large models face a dilemma in training stability, where traditional single-channel connections lead to information congestion as model size increases [3][5]. - Previous solutions, like the hyper-connection approach, improved efficiency but introduced new issues such as uncontrolled information amplification or suppression, leading to gradient explosion and training failures [5][7][9]. Group 2: mHC Architecture - The mHC architecture incorporates an intelligent scheduling system for multi-channel connections, utilizing the Sinkhorn-Knopp algorithm to maintain energy conservation during information transmission [11][13]. - Additional design features include non-negative constraints on input-output mappings to prevent useful signal loss due to coefficient cancellation [15]. Group 3: Infrastructure Optimization - DeepSeek has optimized its infrastructure by merging multiple computation steps into a single operator, reducing memory read/write cycles and employing recomputation strategies to lower memory usage [16][18]. - These optimizations have resulted in significant stability improvements with minimal increases in training time, even at an expansion factor of 4 [18]. Group 4: Performance Validation - Testing on various model sizes, particularly a 27 billion parameter model, demonstrated that mHC effectively resolved training instability issues, achieving lower loss values compared to traditional baseline models [21][22]. - The performance advantages of mHC were consistent across different model sizes, indicating its practical value for both small and large models [24]. Group 5: Industry Implications - The introduction of mHC suggests a shift in the industry towards refined architectural designs rather than merely increasing parameters and computational power, potentially lowering entry barriers for smaller companies in the large-scale model domain [26][29]. - This pragmatic technological innovation is expected to facilitate the deployment of AI technologies, making it easier for more enterprises to engage in large-scale model development [29].

人民财讯1月7日电，工信部印发《工业互联网和人工智能融合赋能行动方案》。其中提出，加快技术产 品创新。鼓励人工智能企业、工业互联网企业、工控企业联合推进工业通信芯片、工业传感器、工业终 端、工业控制系统等智能化升级，逐步深化人形机器人应用。利用人工智能优化工业软件开发流程和模 式，提升主动优化、辅助生成等能力，降低开发门槛和成本。研发智能生产调度管理软件、工艺参数自 优化软件等通用工具产品。推动人工智能赋能网络安全技术创新。加强工业互联网与人工智能标准体系 衔接，完善与人工智能技术相融合的工业互联网体系架构，加强标准宣贯和应用推广。 ...

智通财经APP获悉，近日，备受关注的智谱AI港股招股收官，并将于1月8日以"2513"为股票代码，正式 在港交所主办挂牌上市。届时，"全球大模型第一股"将正式诞生，估值有望达到511亿港元。同日，另 一家AI大模型企业MiniMax计划赴港IPO定价，拟以每股165港元的价格发行新股，且认购需求强劲。 据悉，智谱AI成立于2019年，是"大模型六小龙"中最早成立的企业，按2024年收入计，其在国内独立通 用大模型开发商中位列第一。MiniMax是全球领先的通用人工智能科技公司，旗下已推出MiniMax、 MiniMax语音、海螺AI等产品。2025年前9个月，公司营收同比增长超过170%。 两大AI企业火热登陆港股，再度引爆AI赛道。根据UNCTAD的报告，全球AI市场规模预计将从2023年 的1890亿美元飙升至2033年的4.8万亿美元。作为AI应用层的典型代表，AIGC软件A股上市公司万兴科 技（300624.SZ）等AI头部企业，也再度受到市场关注。 从AI产业层面看，大模型正推进AI企业的发展分化。以智谱AI为代表的通用大模型厂商，核心在于夯 实模型底座，通过API调用向开发者和企业输出通用智能能力 ...