GLM-4.7 主打编程与代理式任务（coding + agentic tasks）的提升 ，同时在 推理能力 等方面也有所增强。 众所周知，今年的大模型，已经从卷"会答题"发展成了 卷"能干活" 。对应的变化，也体现在评测体系上，这些测试覆盖了真实代码修改、终端操作、多 工具调用以及长链路任务执行等场景。 作者 ｜ 木子、高允毅 正在 冲刺大模型第一股 的 智谱 ，今天又拿出了诚意满满的新一代大模型 GLM-4.7，还给开源了 。 智谱也是 把 GLM-4.7 一口气送进了 17 项基准测试 ，和 GPT-5、Claude Sonnet 4.5、Gemini 3.0 Pro、DeepSeek-V3.2、Kimi K2 Thinking 等同台竞 技。 然后在一众强劲对手中，GLM-4.7 还在 两项 BenchMark 中刷新了公开 SOTA ：在 AIME 2025 （测试高级数学推理能力）中正确率已达 95.7%；在 BrowseComp-ZH 中分数达 66.6%。 | Benchmark 基准 | GLM-4.7 | GLM-4.6 | Kimi K2 Thinking | DeepSeek- V3 ...

Core Insights - The article emphasizes that 2025 will be a pivotal year for Agents to transition from a technical concept to mainstream commercial use, making it essential for both businesses and individuals to adopt Agents for survival in the era of intelligence [2]. What is an Agent? - An Agent is defined as an "autonomous intelligent entity" capable of perceiving its environment, analyzing objectives, making decisions, and continuously evolving. Unlike traditional AI, which is viewed as a tool, Agents function more like digital assistants [2]. - For programmers, Agents are not merely "chatbots" but "super plugins" that can autonomously break down tasks using frameworks like LLM and reinforcement learning [2]. How to Embrace Agents? - To help developers quickly understand the core technologies behind Agents, a recommended resource is a two-hour video course titled "Agent Development Methodology in the Era of Large Models," created by Peng Jing Tian, a Google Developer Expert [4]. Cognitive Upgrade and Skill Reconstruction - The article suggests a shift in mindset from focusing on "AI replacing jobs" to considering "how to leverage Agents to amplify personal value" [6]. - It highlights the importance of mastering new collaborative languages for working with Agents, such as prompt engineering, goal decomposition, and human-machine collaboration [6]. Additional Resources - The article mentions a series of supplementary learning materials, including a "China AI Agent Product Compass," an "AI Agent Industry Research Report," and course-related materials to provide a more systematic understanding of Agents [8]. - A knowledge base on Agents is also available, offering insights into various frameworks and applications [10]. Industry Applications - The article notes that Agents are gaining traction due to their ability to execute tasks autonomously and their broad applicability across various industries, including healthcare, education, and finance [20].

前段时间，Anthropic 正式发布了 Claude Sonnet 4.5，对它的定位是"世界上最好的编码模型"和"构建复杂智能体的最强模型"。 凭什么这么说呢，Anthropic 在客户测试中观察到，Claude 4.5 能连续专注干活超过 30 小时，而上一代的数据是 7 个小时。以前是代替 1 个程序员， 现在可以代替 4 个了呢。 更夸张的是，它能连续敲出约 11,000 行代码，快速搞出来一款聊天应用。我费劲工作一个月还不如 AI 轻松工作一天。好好好，既生我，何生 AI …… | | Claude | Claude | Claude | GPT-5 | Gemini | | --- | --- | --- | --- | --- | --- | | | Sonnet 4.5 | Opus 4.1 | Sonnet 4 | | 2.5 Pro | | | 77.2% | 74.5% | 72.7% | 72.8% GPT-5 | | | Agentic coding SWE-bench Verified | 82.0% | 79.4% | 80.2% | | 67.2% | | | with pa ...

Core Viewpoint - The launch of Doubao Mobile Assistant by ByteDance signifies a significant shift in the application paradigm of large models, transitioning from "Chat" to "Action," establishing it as the first system-level GUI Agent in the industry [2][3]. Technical Analysis and Evaluation - The core technology of Doubao Mobile Assistant is the GUI Agent, which has evolved from an "external framework" to a "model-native intelligent agent" between 2023 and 2025. The early stage (2023-2024) relied on external frameworks that limited the agent's capabilities due to dependency on prompt engineering and external tools [4]. - The introduction of visual language models driven by imitation learning in 2024 marked a shift to model-native capabilities, allowing the agent to understand interfaces directly from pixel inputs, significantly enhancing adaptability to unstructured GUIs [5]. - By 2024-2025, reinforcement learning-driven visual language models became mainstream, enabling agents to autonomously execute tasks in dynamic environments. Doubao Mobile Assistant embodies this technological evolution [5][7]. Development History of GUI Agent - Previous GUI Agents were often limited to demo stages due to reliance on Android accessibility services, which had significant drawbacks. Doubao Mobile Assistant overcomes these issues through a customized OS that allows for non-intrusive system-level control [7][8]. - The model architecture of Doubao Mobile Assistant employs a collaborative end-cloud model, indicating a shift from experimental to practical applications of GUI Agents [8]. Limitations and Future Outlook - Doubao Mobile Assistant faces three major challenges: security risks associated with cloud-side model reliance, insufficient autonomous task completion capabilities, and limited ecological coverage [9][10][11]. - The assistant currently operates as a passive tool, lacking personalized proactive service capabilities. Future developments must focus on enhancing privacy, environmental perception, complex decision-making, and personalized service [12][13]. Evolution of End-Side Intelligence - The emergence of system-level GUI Agents presents a fundamental contradiction between the need for comprehensive operational visibility and user privacy concerns. A balance must be struck to ensure user data sovereignty while providing intelligent services [13][14]. - The future AI mobile ecosystem should adhere to the principle of "end-side native, cloud collaboration," ensuring that sensitive user data remains on-device while leveraging cloud capabilities for complex tasks [14][15]. Autonomous Intelligence and User Interaction - Doubao Mobile Assistant's current capabilities are based on extensive data training, but future autonomous intelligence must enable agents to learn and adapt in dynamic environments, overcoming challenges in generalization, autonomy, and long-term interaction [22][24][25]. - The transition from passive execution to proactive service is essential for personal assistants to reduce user cognitive load and enhance user experience [29][30][31]. Industry Trends and Future Predictions - In the short term (within one year), more mobile assistants are expected to launch, intensifying competition between application developers and hardware manufacturers [35]. - In the medium term (2-3 years), the concept of a "personal exclusive assistant" will solidify, with end-side models evolving to provide personalized experiences based on user data [36]. - In the long term (3-5 years), a new type of end-side hardware will emerge, integrating high privacy operations and lightweight tasks, ensuring data sovereignty and rapid response times [38].

Core Viewpoint - The article discusses the recent developments of MiniMax, a Chinese AI company, as it approaches its IPO, highlighting its rapid growth and positioning in the general artificial intelligence (AGI) sector amidst a changing capital market narrative [2][15]. Group 1: Company Overview - MiniMax, founded in early 2022, is recognized as one of the first companies in China to apply the "Mixture of Experts" (MoE) architecture on a large scale [4]. - The company aims to develop globally competitive AGI models and has a cash balance of approximately $1.046 billion as of September 30, 2025, ensuring sufficient funding for ongoing model development and computational resources [4]. - MiniMax has a workforce of 385 employees with an average age of 29, indicating a youthful and dynamic leadership team [4]. Group 2: Product Development - MiniMax plans to release the abab 6.5 model in the first half of 2024, which is expected to be one of the first commercialized trillion-parameter MoE models in China [8]. - The company has developed a multi-modal model system covering text, speech, images, and video, with capabilities being modularized for unified output to products and APIs [8]. - MiniMax has launched several AI-native products, including "海螺 AI" for multi-modal content creation and "星野" for virtual companionship, which have gained traction in both domestic and international markets [9]. Group 3: Financial Performance - MiniMax's R&D expenditures from 2022 to 2025 are projected to total approximately $500 million, significantly lower than OpenAI's estimated $40 billion to $55 billion, showcasing a high efficiency in investment [10]. - The company's gross margin improved from -24.7% in 2023 to 12.2% in 2024, and further to 23.3% in the first nine months of 2025, attributed to revenue growth and operational efficiency [11]. - Revenue has shown a rapid increase, with figures of $0 in 2022, $3.46 million in 2023, and $30.52 million in 2024, with a projected revenue of approximately $53.44 million for the first nine months of 2025, indicating a year-on-year growth of over 170% [12]. Group 4: Market Position and Capitalization - MiniMax has attracted significant investment, completing seven funding rounds with early investors including prominent firms like Tencent and Alibaba, achieving a valuation of over $4.2 billion after its latest funding round [14]. - The company has expanded its paid user base from approximately 119,700 in 2023 to about 1.77 million by September 30, 2025, serving over 200 million individual users globally [14]. - The IPO process for MiniMax is seen as a signal for the broader industry, pushing companies to focus on quantifiable metrics and operational narratives rather than just technological capabilities [15][16].

Group 1 - The core viewpoint of the article emphasizes the rapid advancements in AI, particularly in code generation, while also highlighting the associated risks and challenges, as noted by Sergey Brin [2][3][20] - Brin pointed out that AI's ability to write code can lead to significant errors, making it more suitable for creative tasks where mistakes are less critical [2][38] - He reflected on Google's initial hesitations regarding generative AI and the underestimation of the importance of scaling computational power and algorithms [2][22][24] Group 2 - The discussion included a historical overview of Google's founding, emphasizing the creative and experimental environment at Stanford that fostered innovation [4][6][10] - Brin noted that the early days of Google were characterized by a lack of clear direction, with many ideas being tested without strict limitations [6][9] - The importance of a strong academic foundation in shaping Google's culture and approach to research and development was highlighted [12][13] Group 3 - Brin discussed the competitive landscape of AI, noting that significant investments in AI infrastructure have reached hundreds of billions, with companies racing to lead in this space [21][22] - He acknowledged that while Google has made substantial contributions to AI, there were missed opportunities in the past due to insufficient investment and fear of releasing products prematurely [22][23][24] - The conversation also touched on the evolving nature of AI, with Brin expressing uncertainty about its future capabilities and the potential for AI to surpass human abilities [27][29][30] Group 4 - Brin emphasized the need for a balance between computational power and algorithmic advancements, stating that algorithmic progress has outpaced scaling efforts in recent years [3][55] - He mentioned that deep technology and foundational research are crucial for maintaining a competitive edge in AI [24][25] - The discussion concluded with reflections on the role of universities in the future, considering the rapid changes in education and knowledge dissemination due to technology [41][42]

Group 1 - Alibaba faced rumors about a fake all-hands meeting involving employees holding dumplings, which the company quickly denied, stating the images were AI-generated [3][9][10] - Zhou Hongyi, founder of 360, was accused of financial fraud by a former executive, claiming to have evidence of at least tens of billions in false accounting [13][16] - Tencent has restructured its AI development framework, appointing renowned AI researcher Yao Shunyu as Chief AI Scientist to enhance its capabilities [18][19] Group 2 - ByteDance announced significant salary increases and performance bonuses to attract and retain talent, with a 35% increase in the performance evaluation cycle for 2025 [21][23] - The company is also collaborating with hardware manufacturers like Vivo and Lenovo to develop AI smartphones, aiming to create new monetization pathways [24][26] - Beijing Zhiyu Huazhang Technology has submitted its IPO application, reporting substantial R&D investments and a high growth rate in revenue [27][28] Group 3 - Moore Threads unveiled a new GPU architecture capable of supporting large-scale clusters, enhancing performance for AI and graphics applications [29][30] - Musk's 2018 compensation agreement with Tesla, valued at $56 billion, was reinstated by a Delaware court, allowing him to benefit from stock options [31][32] - TikTok's new U.S. strategy involves forming a joint venture for data security while retaining control over its e-commerce and advertising operations [33] Group 4 - An Indian semiconductor company's stock skyrocketed by 550 times in 20 months despite having only two employees and no operational revenue, raising concerns about market speculation [36] - A group was arrested for spreading negative information about brands like Xiaomi and Huawei, highlighting organized efforts to manipulate public perception [37] - Google has been rehiring former employees, particularly in AI roles, indicating a shift in talent acquisition strategy following significant layoffs [38] Group 5 - Manus reported achieving an annual recurring revenue (ARR) of $100 million, with a monthly growth rate exceeding 20% since the release of its latest version [39][40] - Cambricon plans to use nearly 2.8 billion yuan in capital reserves to offset losses, reporting a significant turnaround in profitability for the first three quarters of the year [42][43] - SpaceX is in the process of selecting investment banks for a potential IPO, aiming to capitalize on favorable market conditions [44]

Core Viewpoint - Yann LeCun criticizes the current AI development path focused on scaling large language models, arguing it leads to a dead end and emphasizes the need for a different approach centered on understanding and predicting the world through "world models" [2][3]. Group 1: AI Development Path - LeCun believes the key limitation in AI progress is not reaching "human-level intelligence" but rather achieving "dog-level intelligence," which challenges the current evaluation systems focused on language capabilities [3]. - He is establishing a new company, AMI, to pursue a technology route that builds models capable of understanding and predicting the world, moving away from the mainstream focus on generating outputs at the pixel or text level [3][9]. - The current industry trend prioritizes computational power, data, and parameter scale, while LeCun aims to redefine the technical path to general AI by focusing on cognitive and perceptual fundamentals [3][9]. Group 2: Research and Open Science - LeCun emphasizes the importance of open research, stating that true research requires public dissemination of results to ensure rigorous methodologies and reliable outcomes [7][8]. - He argues that without allowing researchers to publish their work, the quality of research diminishes, leading to a focus on short-term impacts rather than meaningful advancements [7][8]. Group 3: World Models and Planning - AMI aims to develop products based on world models and planning technologies, asserting that current large language model architectures are inadequate for creating reliable intelligent systems [9][10]. - LeCun highlights that world models differ from large language models, as they are designed to handle high-dimensional, continuous, and noisy data, which LLMs struggle with [10][11]. - The core idea of world models is to learn an abstract representation space that filters out unpredictable details, allowing for more accurate predictions [11][12]. Group 4: Data and Learning - LeCun discusses the vast amount of data required to train effective large language models, noting that a typical model pre-training scale is around 30 trillion tokens, equating to approximately 100 trillion bytes of data [20]. - In contrast, video data, which is richer and more structured than text, offers greater learning value, as it allows for self-supervised learning due to its inherent redundancy [21][28]. Group 5: Future of AI and General Intelligence - LeCun expresses skepticism about the concept of "general intelligence," arguing it is a flawed notion based on human intelligence, which is highly specialized [33][34]. - He predicts that significant advancements in world models and planning capabilities could occur within the next 5 to 10 years, potentially leading to systems that approach "dog-level intelligence" [35][36]. - The most challenging aspect of AI development is achieving "dog-level intelligence," after which many core elements will be in place for further advancements [37]. Group 6: Safety and Ethical Considerations - LeCun acknowledges the concerns surrounding AI safety, advocating for a design approach that incorporates safety constraints from the outset rather than relying on post-hoc adjustments [43]. - He argues that AI systems should be built with inherent safety features, ensuring they cannot cause harm while optimizing for their objectives [43][44].

作者 | 褚杏娟 所以，这就是我们的初衷：如果我们设计专门用于这类机器学习计算的硬件，也就是密集低精度线性代数相关的硬件，就能大 幅提升效率。事实也证明了这一点。第一代 TPU 的能效比当时的 CPU 或 GPU 高出 30 到 70 倍，速度也快 15 到 30 倍。 根据最新报道，随着谷歌 TPU 芯片需求大涨，谷歌扩大了对联发科合作定制新一代 TPU v7e 的订单，订单量比原规划激增数 倍。消息称，联发科为谷歌操刀定制的首款 TPU v7e 将于下季度末进入风险性试产，并再拿下谷歌下一代 TPU v8e 的订单。 联发科大单获得了台积电的先进封装产能支持，2027 年台积电提供给联发科谷歌项目的 CoWoS 产能更将暴增 7 倍以上。 尽管承认谷歌在过去 10 年中取得了进步，但英伟达认为其大约领先谷歌 TPU 两年。由于人工智能模型变化迅速，英伟达认为 谷歌很难让云服务提供商采用 TPU，因为 TPU 是为更特定的模型类型而设计的。相比之下，英伟达相信其更灵活、可编程的 平台仍然是构建大规模云端人工智能基础设施的最佳选择。 但无论如何，谷歌确实让英伟达产生了些许危机。近日，在 NeurIPS 大会期 ...

Core Insights - OpenAI's CEO Sam Altman expresses concerns about competition, particularly from Google, which he views as a significant threat to OpenAI's market position [2][11] - Altman emphasizes the importance of user retention and the development of "AI-native software" rather than merely integrating AI into existing products [2][12] - OpenAI is focusing on creating a comprehensive product ecosystem that enhances user experience through personalization and memory capabilities [9][10] Group 1: Competition and Market Position - Altman acknowledges that OpenAI is in a "red alert" state due to increasing competition, particularly after the release of Google's Gemini 3, but believes the impact has not been as severe as initially feared [5][6] - He notes that while Google has a strong distribution advantage, OpenAI's user base has grown significantly, reaching nearly 9 million users, which provides a competitive edge [3][8] - Altman believes that maintaining a slight paranoia about competition is beneficial for OpenAI's strategy and product development [6][7] Group 2: Product Development and Strategy - OpenAI is not rushing to release GPT-6; instead, it plans to focus on customized upgrades that cater to specific user needs, with significant improvements expected in early 2024 [36][37] - The company aims to build the best models and products while ensuring sufficient infrastructure to support large-scale services [8][9] - Altman highlights the importance of creating a cohesive product ecosystem that integrates various functionalities, making it easier for users to adopt and rely on OpenAI's offerings [10][24] Group 3: Enterprise Market Focus - OpenAI's strategy has shifted towards prioritizing enterprise solutions, as the technology has matured enough to meet business needs [27][28] - The company has seen rapid growth in its enterprise segment, with increasing demand for AI platforms from businesses [28][29] - Altman emphasizes that the enterprise market is ready for AI integration, particularly in areas like finance and customer support [29][30] Group 4: Infrastructure and Financial Outlook - OpenAI has committed approximately $1.4 trillion to build its infrastructure, which is essential for supporting its AI capabilities and future growth [39][48] - The company anticipates that as revenue grows, the cost of inference will eventually surpass training costs, leading to profitability [48][49] - Altman acknowledges that while current spending is high, the long-term vision is to create a sustainable business model that leverages AI advancements [50][51]