Group 1: Event Overview - The 2025 AI Partner Conference, co-hosted by 36Kr and CEIBS, was held in Beijing, focusing on "Chinese Solutions" and the future of AI [1] - The conference featured discussions on four main topics: the golden moment of Chinese innovation in AI, the potential of superintelligent agents, the reshaping of global tech competition by Chinese solutions, and the integration of AI across various industries [1] Group 2: Company Insights - The company, established in May 2023, has rapidly grown by leveraging diverse and collaborative core technologies, partnering with nearly 100 entities across AI models, chips, and industry clients [3] - The company aims to democratize AI through technological innovation, likening computational power to the foundational resources of water and electricity in the industrial era [3] Group 3: Challenges and Solutions - The journey towards AGI (Artificial General Intelligence) faces a core contradiction: the need for expanding computational resources to meet infinite demands, which may hinder AGI development due to resource limitations [4] - The proposed "dual approach" solution includes enhancing resource utilization efficiency and expanding the scale of computational resources to lay the groundwork for the AGI era [4][5] Group 4: Product Offerings - The company has introduced a product system comprising "large, medium, and small boxes" to address varying computational needs [6] - Large Box: "Wuqing AI Cloud" for large-scale computational demands, integrating resources from 26 provinces and 53 data centers, capable of supporting over 25,000 P of computational power [6] - Medium Box: "Wujie Intelligent Computing Platform" focuses on activating domestic computational resources and providing tailored intelligent computing services [6] - Small Box: Solutions for edge computing devices, optimizing efficiency for AI applications on terminals like smartphones and PCs [6] Group 5: Ecosystem and Collaboration - The "Wuqing AI Cloud" supports a standardized and open interface, facilitating a unique "platform + self-operated" model that promotes collaborative innovation across the industry [7] - The company has achieved significant milestones, such as surpassing an average daily token call volume of 10 billion on its platform, supporting over 100 AI applications [7][8] Group 6: Industry Applications - The company's products have been successfully applied in various scenarios, including AIGC (AI-Generated Content) and AI recruitment, providing comprehensive services to enhance user experience and operational efficiency [8][9] - The "Wujie Intelligent Computing Platform" has enabled significant advancements in AI model training and deployment, achieving notable results in collaboration with research institutions and industry partners [9][10] Group 7: Future Outlook - The company aims to empower various industries through AI technology, emphasizing the vast market potential and the early stage of industry development [10]

Core Viewpoint - The article discusses the challenges and opportunities in China's computing power market, particularly focusing on the emergence of "intelligent computing centers" and the role of the company Wuwen Xinqun in addressing these challenges through innovative solutions [1][10]. Group 1: Current State of Computing Power - As of September 2024, China's computing power scale reached 246 EFLOPS, with intelligent computing power growing over 65% year-on-year, and over 13,000 computing power application projects across various industries [1]. - Despite the growth in computing power, the average cabinet utilization rate in intelligent computing centers is only 20% to 30%, with some enterprise-level centers as low as 10% [1][2]. - The overall utilization rate of computing power in the country is only 32%, indicating a significant gap between supply and demand [2]. Group 2: Challenges in the Computing Power Market - There is a shortage of quality computing power supply, making it difficult for many enterprises to find suitable resources [2][3]. - High usage thresholds prevent startups from effectively utilizing available computing power, leading to a phenomenon of "computing power scarcity" among AI companies [3]. - The domestic chip ecosystem is fragmented, with various models and architectures that are incompatible, hindering efficient resource flow [3]. Group 3: Wuwen Xinqun's Approach - Founded in May 2023 by a team with strong ties to Tsinghua University, Wuwen Xinqun has quickly gained market attention, securing nearly 1 billion yuan in funding within two years [4][5]. - The company aims to be a "computing power operator" in the era of large models, addressing the challenges posed by the dominance of NVIDIA's CUDA ecosystem and the fragmentation of domestic chip manufacturers [5]. - Wuwen Xinqun has developed a cloud computing network that integrates heterogeneous computing resources, allowing developers to focus on applications without worrying about underlying hardware differences [5][10]. Group 4: Product Offerings - Wuwen Xinqun has introduced three core products, referred to as "three boxes," to enhance computing power utilization across various scales [6][9]. - The "Wuqiong AI Cloud" serves as a large-scale computing network, integrating resources from 26 provinces and over 53 data centers, with a total computing power exceeding 25,000 P [7]. - The "Wujie Intelligent Computing Platform" targets large computing clusters, demonstrating significant performance in training large models [8]. - The "Wuyin Terminal Intelligence" solution is designed for limited computing terminals, enabling them to perform complex tasks without relying on cloud resources [9]. Group 5: Future Outlook - Wuwen Xinqun's efforts reflect a broader need for a cohesive ecosystem that allows computing power to flow effectively, addressing the current challenges of fragmentation and high costs [10][11]. - The company aims to enhance the actual utilization rate of computing resources and improve the cost-performance ratio, contributing to the growth of China's AI industry [10].

Core Insights - The company, Wunwen Xinqiong, launched a comprehensive AI efficiency enhancement solution during the WAIC 2025, introducing three core products: Wunqiong AI Cloud, Wujie Intelligent Computing Platform, and Wuyin Terminal Intelligence [2][3] - The Wunqiong AI Cloud targets global computing networks ranging from 10,000 to 100,000 cards, while the Wujie Intelligent Computing Platform is designed for large computing clusters of 100 to 1,000 cards, and the Wuyin Terminal Intelligence focuses on limited computing terminals from 1 to 10 cards [2][3] - The company aims to create a "universal language" for the industry to enable seamless communication and collaboration between different chip architectures, thereby enhancing the efficiency of AI technology deployment [3][4] Product Offerings - Wunwen Xinqiong's product ecosystem spans from infrastructure to industry applications, supporting AGI technology's large-scale implementation [6] - The company reported that its AI cloud platform provides a one-stop service for enterprises and developers, including cloud management, foundational cloud products, and large model development platforms [5][6] - The company has achieved significant performance optimization for over ten types of domestic AI chips, enhancing their performance by 50% to 200% through algorithm and compilation optimizations [5][6] Market Position and Growth - Wunwen Xinqiong completed nearly 500 million yuan in Series A financing in 2024, setting a record for the largest single financing in domestic AI infrastructure [5] - The company has raised over 1 billion yuan since its establishment in May 2023, indicating strong investor confidence and market potential [5] - The company serves the world's largest AI incubator, Shanghai Mosu Space, which has surpassed 10 billion daily token calls, supporting over 100 innovative AI applications [6] Strategic Vision - The CEO emphasized the importance of creating a closed-loop ecosystem for chip and model development to guide future advancements in the industry [4][8] - The company aims to achieve its ultimate vision of "limitless intelligence and precise computing" by harmonizing scene scale, computing resources, and intelligent efficiency [9]

Group 1 - Anthropic announced a weekly usage limit for Claude Pro and Max users, affecting less than 5% of subscribers [1] - Some users reported extreme cases where a $200 plan resulted in actual consumption of tens of thousands of dollars due to continuous operation [1] - Users expressed a lack of transparency regarding usage, leading many to seek alternative products [1] Group 2 - Microsoft Edge introduced a "Copilot mode" that enhances context awareness across tabs, allowing simultaneous reading and analysis of all open pages [2] - The new interface features a simplified input box that understands user intent and supports voice control and thematic journey functions [2] - This feature is currently available for free in all Copilot markets but may be bundled with a subscription service in the future [2] Group 3 - Wuwen Chipong launched a comprehensive AI efficiency enhancement solution, including three core products: Wuqiong AI Cloud, Wujie Intelligent Computing Platform, and Wuyin Terminal Intelligence [3] - The solution covers 26 provinces and cities with 53 core data centers, integrating over 15 mainstream chip architectures and achieving a total computing power scale exceeding 25,000 P [3] - Innovations on the edge include the world's first edge intrinsic model "Wuqiong Tianquan," which maintains cloud-level intelligence with 21 billion parameters while controlling memory usage to 7 billion [3] Group 4 - Step 3 launched a new AI research assistant called "Jieyue Deep Research," capable of completing complex research tasks and generating in-depth professional reports within ten minutes [4][5] - The assistant achieved a 70% high pass rate in the xbench-DeepSearch evaluation [5] - It is based on reinforcement learning and multi-agent architecture, enabling autonomous thinking, reasoning, and dynamic tool usage for real-world complex tasks [5] Group 5 - JD.com upgraded its large model brand to JoyAI, introducing solutions like JoyAgent intelligent agent platform, JoyInside embedded intelligence, and digital humans [6] - JoyAgent is the first 100% open-source enterprise-level intelligent agent, receiving over 2,000 GitHub stars and possessing a complete product-level closed-loop capability [6] - JoyAI's products have been implemented in various scenarios, with digital human services exceeding 20,000 brands and the interactive AI toy Fuzozo selling out during its first pre-sale [6] Group 6 - Researchers from UC San Diego and NYU launched and open-sourced MIRIX, the world's first multi-modal, multi-agent AI memory system, along with a desktop app [7] - The system categorizes memory into six modules: core, context, semantics, programs, resources, and knowledge repository, managed by a meta-memory manager and six memory sub-modules [7] - MIRIX achieved a 35% higher accuracy than traditional RAG in the ScreenshotVQA test and reduced storage by 99.9%, setting a record of 85.4% in the LOCOMO long dialogue task [7] Group 7 - The National Satellite Meteorological Center, Nanchang University, and Huawei jointly released the "Fengyu" model, the world's first full-chain space weather AI forecasting model [8] - The model features a pioneering chain training structure, including solar wind, Earth's magnetic field, and ionosphere models [8] - In practical tests, "Fengyu" maintained a prediction error of around 10% for global electron density and performed excellently during multiple major magnetic storm events, with 11 national invention patents applied [8] Group 8 - Shanghai AI Lab released and open-sourced the "Shusheng" scientific multi-modal large model Intern-S1, which surpasses top closed-source models in scientific capabilities [9] - The model features a "cross-modal scientific analysis engine" that can accurately interpret complex scientific data such as chemical formulas and protein structures [9] - The research team proposed a method for synthesizing scientific data that combines general reasoning capabilities with multiple top professional abilities, creatively reducing reinforcement learning training costs [9] Group 9 - a16z partner Martin Casado stated that the AI large model competition will evolve into an oligopoly similar to the cloud computing battle, creating a new brand effect [10] - In AI competition, the application layer lacks a technological moat, and rational business decisions will focus on "sacrificing profits for distribution," with value emerging from foundational infrastructure and vertical domain deepening [10] - AI will not transform ordinary developers into super engineers but will allow "10x engineers to become 2x," simplifying programming by eliminating cumbersome tasks and returning to the essence of creation [10] Group 10 - Tencent's Robotics X Lab and Futian Lab jointly launched the embodied intelligence open platform Tairos, aimed at enhancing software capabilities for robot developers and application developers [11] - The platform is based on the SLAP³ technology system, providing three core capabilities: planning large models, multi-modal perception large models, and perception-action joint large models [11] - Five major trends in the future development of embodied intelligence were identified: integration of virtual and real worlds, reduced technical barriers, intelligent evolution, agentification, and multi-modal perception [11]

Core Viewpoint - The demand for AI computing power is surging, leading to a focus on the diversification and localization of computing resources in the industry [1] Group 1: AI Computing Power Landscape - The current domestic chip and computing resource landscape is diverse, with multiple independent ecosystems, but significant differences in hardware architecture and interface protocols hinder the efficiency of AI technology implementation [1] - The company, Wunwen Xinqiong, has developed a "universal language" for the industry that enables seamless communication and collaboration between different chips, allowing developers to avoid the complexities of varying chip usage [1][2] - The rise of domestic computing power is expanding the available resource pool, with domestic computing power now accounting for nearly half of the overall deployment, particularly excelling in inference and certain training scenarios [7] Group 2: Business Model and Services - Wunwen Xinqiong's business model is characterized by a "universal" approach, aiming to provide precise support for small and medium-sized enterprises during their growth phases through flexible computing power services [6][8] - The company offers various billing methods for its services, including per card, per hour, or based on usage volume, which transforms fragmented computing resources into standardized services [8] - The product matrix includes "large box," "medium box," and "small box" offerings, focusing on national-level computing power scheduling and integrating computing service capabilities into AI clusters and terminal devices [9] Group 3: Industry Trends and Future Outlook - The AI computing power industry is transitioning from a "technical dividend period" to a "value closed-loop period," with the core issue shifting from whether AI is usable to whether it is worth using, making cost-effectiveness a critical breakthrough point [10] - The company is exploring the customization of high-performance, cost-effective dedicated chips for edge AI applications, driven by the vast user base and manufacturing capabilities in China [10][11] - The construction of an ecosystem that connects models, systems, and hardware is a long-term goal, with the company aiming to create a positive cycle of hardware iteration, model optimization, and scene implementation [11]

衡宇 发自 WAIC 量子位 | 公众号 QbitAI 智能时代的尺度，在计算资源与智能效率的双重牵引下正在极速压缩、迅速蔓延。 两年前，我们惊艳于几千卡集群训练而成的GPT3.5；但今天，一部手机也可以装下与它同等性能的小型AI了。 2025年WAIC上， 无问芯穹联合创始人、CEO夏立雪 如此说道。 他还代表无问芯穹，带来了AI落地这道难题的最新回答—— 三个盒子，打通从十万卡到一张卡的AI效能跃升路径 。 是的，仅仅是三个盒子。 在无问芯穹看来，这三个盒子背后，是一整套面向未来的智能基础设施设计。 什么是三个盒子？ "三个盒子"其实是无问芯穹全规模AI效能跃升方案的三大核心产品： 这是一整套软硬件协同系统，专为未来智能基础设施设计，能覆盖从云到端的各种规模场景，支持多种异构算力，同时打通模型调度、性能优 化到应用部署的全流程。 我们一个一个来看—— 大盒子：无穹AI云 大盒子：无穹AI云 中盒子：无界智算平台 小盒子：无垠终端智能 大盒子，即无问芯穹推出的 无穹AI云 ，是面向万卡至十万卡级别的智算网络，为超大规模算力集群的利用提供了一个系统性的解决方案。 夏立雪在现场透露，无界智算平台已在超过100个 ...

整体来讲，这一 方案是一套面向未来智能基础设施的软硬协同系统，为跨地域智算网络、智算集群 与多形态智能终端等全规模场景，统一适配多种异构算力，提供从模型调度、性能优化到应用部署的 全链路支持。 夏立雪表示，无问芯穹希望通过提供 "打包式"的产品服务能力，在单卡至十万卡算力的全规模软硬 件场景中，让每一份算力，都能释放最大的智慧潜能。 夏立雪指出，从传统算法，到 AI1.0、AI2.0阶段，在Scaling Law的推动下，计算资源持续驱动着 智能边界的拓展，逼近AGI的临界点。然而，有一条人类文明的终极边界始终横亘在AGI之路上—— 资源的有限性。 人类文明，在迎来一个 "无所不能"的智慧之前，或将首先触碰到资源总量的红线。 无问芯穹发布全规模AI效能跃升方案，要以有限的资源实现"无限"的需求。 本文为IPO早知道原创 作者｜ Stone Jin 微信公众号｜ipozaozhidao 据 IPO早知道消息， 无问芯穹联合创始人、 CEO夏立雪 于 7月28日发 布了无问芯穹全规模 AI效 能跃升方案，并正式推出三大核心产品：针对万卡至十万卡全局算力网络的"无穹AI云"、针对百卡 至千卡级大型智算集群的"无界智 ...

Group 1 - The 2025 World Artificial Intelligence Conference showcased unprecedented scale, with over 800 companies and more than 3,000 cutting-edge exhibits, including 40 large models and 60 smart robots [2][5] - Baidu introduced its new digital human technology, NOVA, which aims to replicate top-tier influencer capabilities and is expected to be available to the industry by October [5][6] - The AI-powered educational tool, Youdao AI Answering Pen SpaceOne, integrates advanced AI models to assist students in problem-solving across various subjects [6][8] Group 2 - Black Lake Technology presented its AI Agent-based flexible manufacturing solution, which enhances order response speed and fulfillment efficiency for small and medium-sized factories [11][12] - SenseTime launched its embodied intelligence platform "Wuneng," designed to provide robots and smart devices with advanced perception and interaction capabilities [12][13] - Wuneng aims to facilitate interaction between intelligent devices and the real world, enhancing operational efficiency in various applications [12][13] Group 3 - The conference emphasized the importance of international collaboration in AI governance to ensure the development of "good AI" that does not threaten human existence [15][17] - Discussions highlighted the need for equitable access to AI technology and the importance of addressing safety concerns to benefit all nations [17][18] - Experts called for effective regulation of AGI to minimize risks and ensure it serves as a global public resource [18]

Core Insights - The future of "smart terminals" is being reimagined with the introduction of AIPC (Artificial Intelligence Personal Computer), which can perform tasks while in sleep mode and support complex reasoning locally without relying on cloud computing [1][2][5] Group 1: AIPC and Its Innovations - AIPC is evolving from a simple office tool to a "mobile productivity center," addressing two main pain points: reliance on cloud computing and limited local hardware resources [2] - The "small box" solution, termed "limitless terminal intelligence," aims to maximize the potential of limited terminal resources through software and hardware collaboration [2][5] - The AIPC can release over 1,000 hours of productivity annually by executing background tasks during idle CPU periods, thanks to the Infini-Megrez2.0 model, which achieves cloud-level intelligence with significantly reduced memory and computational requirements [5][7] Group 2: Technological Advancements - The Infini-Megrez2.0 model enables "unperceived operation," transforming AIPC from a mere tool to an intelligent assistant that can autonomously handle tasks during sleep periods [7] - The Infini-Mizar 2.0 inference engine enhances the speed and efficiency of large model operations, increasing the local model size limit from 7B to 30B, while improving intelligence levels by 18% and doubling inference performance [8] - The combination of Mizar2.0 and Megrez2.0 allows for significant advancements in terminal intelligence applications, pushing the boundaries of what is possible in edge AI [8] Group 3: Partnerships and Future Goals - The company is collaborating with Lenovo, H3C, and other partners to turn innovations into accessible products, including an integrated large model machine and FPGA-based inference systems [9][11] - The ultimate goal is to create next-generation smart terminals that serve a wide range of applications, making AGI accessible to everyone [11]

机器之心发布 机器之心编辑部 「 算力是智能时代的土壤，其规模与效率决定着数字未来的疆界。 」 7 月 28 日，2025 年世界人工智能大会上，无问芯穹联合创始人、CEO 夏立雪发布了 无问芯穹全规模 AI 效能跃升方案，并正式推出三大核心产品：无穹 AI 云、无界智算平台与无垠终端智能 。该方案是一套面向未来智能基础设施的软硬协同系统，为跨地域智算网络、智算集群与多形态智能终端等全规模场 景，统一适配多种异构算力，提供从模型调度、性能优化到应用部署的全链路支持。 发布会现场，夏立雪将这三个产品比作了 「 三个盒子 」 ，他表示，无问芯穹希望通过提供 「 打包式 」 的产品服务能力，在单卡至十万卡算力的全规模软 硬件场景中，让每一份算力，都能释放最大的智慧潜能。 1. 两条 「 加速进路 」 和一个 「 价值空间 」 ，让有计算的地方就有智能 夏立雪指出，从传统算法，到 AI1.0、AI2.0 阶段，在 Scaling Law 的推动下，计算资源持续驱动着智能边界的拓展，逼近 AGI 的临界点。然而，有一条 人类文明的终极边界始终横亘在 AGI 之路上 —— 资源的有限性。 人类文明，在迎来一个 「 无所不 ...