以下文章来源于定焦One ，作者定焦One团队 定焦One . 深度影响创新。 本文来自微信公众号： 定焦One ，作者：王璐，编辑：阿伦，题图来自：AI生成 能各不相同，什么是AI浏览器？ 可以理解为将AI融入到传统浏览器中，具备智能搜索与问答、内容理解与生成摘要、自动化任务执 行、多模态交互体验、个性化推荐等功能，实现从"工具"到"智能"的转变。 很多从业者认为，AI浏览器和Agent一样，都属于大势所趋，但此刻大家围绕AI浏览器的谜团也有很 多。 目前各家公司在AI浏览器的技术路线与产品形态上有着明显区别，究竟何种形态才能称之为AI浏览 器？它和Chatbot、Agent在功能和技术上又有何区别？ 先看竞争比较激烈的国外市场。 近期，一则"悬而未决"的竞购在AI圈引发轰动。 靠AI搜索成为行业独角兽的Perplexity称要以345亿美元的价格收购谷歌的Chrome浏览器，这一报价 几乎是Perplexity自身估值 （180亿美元） 的两倍。想买Chrome的还不止Perplexity，另一家AI明星 企业OpenAI，此前也透露出了收购意愿。不过，谷歌方面暂时没有接受任何报价。 看到这个新闻，很多人 ...

证券研究报告 | 行业周报 gszqdatemark 2025 08 24 年 月 日 计算机 第四次谈自定义 Agent 国盛计算机自定义 Agent 相关报告： 2025 年 8 月 10 日——《GPT-5 最受益的方向:自定义 Agent》 2025 年 7 月 27 日——《再谈这轮 AI 持续性》 2025 年 7 月 19 日——《海外 AI 应用的持续性如何》 DeepSeek-V3.1 发布，迈向 Agent 时代的一大步。2025 年 8 月 21 日，DeepSeek-V3.1 正式发布，包含以下主要变化：1）混合推理架构支 持思考模式与非思考模式；2）更高的思考效率：V3.1-Think 在输出 token 数减少 20%-50%的情况下，各项任务的平均表现与 R1-0528 持 平；3）更强的 Agent 能力，新模型在工具使用与智能体任务中的表现有 较大提升；4）上下文已扩展为 128K。5）Base 模型与后训练模型均开 源。DeepSeek-V3.1 报告中还提及使用了 UE8M0 FP8 Scale 的参数精 度，针对即将发布的下一代国产芯片设计，我们认为这意味着 DeepS ...

DeepSeek刚刚官宣V3.1模型。模型同时支持思考模式与非思考模式；思考效率提升，相比DeepSeek-R1-0528，DeepSeek-V3.1-Think能在更短时间内给出答案；Agent能力更强，通过Post-Training优化，新模型在工具使用与智能体任务中的表现有较大提升。 https://t.co/ajVjLgC4riDeepSeek (@deepseek_ai):Introducing DeepSeek-V3.1: our first step toward the agent era! 🚀🧠 Hybrid inference: Think & Non-Think — one model, two modes⚡️ Faster thinking: DeepSeek-V3.1-Think reaches answers in less time vs. DeepSeek-R1-0528🛠️ Stronger agent skills: Post-training boosts tool use and ...

Core Viewpoint - The release of ChatGPT-5 marks a significant advancement in AI capabilities, particularly in reasoning, programming, and creative writing, with notable improvements in reliability and behavior design [3][4][6]. Group 1: Model Improvements - ChatGPT-5 has shown a substantial reduction in issues related to flattery and hallucination, indicating a more reliable interaction model [4][14]. - The model's programming capabilities have seen a qualitative leap, allowing users to create applications with minimal coding knowledge, which is expected to foster the emergence of many small businesses [6][17]. - The team emphasizes the importance of user experience and practical applications as key metrics for evaluating model performance, rather than just benchmark scores [20][21]. Group 2: Training and Development - The development process for ChatGPT-5 involved a focus on desired capabilities, with the team designing assessments to reflect real user value [22][23]. - The integration of deep research capabilities into the model has enhanced its ability to perform complex tasks efficiently, leveraging high-quality data and reinforcement learning [16][26]. - Mid-training techniques have been introduced to update the model's knowledge and improve its performance without the need for extensive retraining [45]. Group 3: Future Implications - The advancements in ChatGPT-5 are expected to unlock new use cases and increase daily usage among a broader audience, which is seen as a critical indicator of progress towards AGI [21][15]. - The model's ability to assist in creative writing has been highlighted, showcasing its potential to help users with complex writing tasks [29][31]. - The future of AI is anticipated to be characterized by the rise of autonomous agents capable of performing real-world tasks, with ongoing research focused on enhancing their capabilities [36][41].

Core Viewpoint - The article discusses the launch of AutoGLM 2.0 by Zhipu, which elevates the application of Agents to a new level, introducing the world's first mobile Agent that operates without occupying user devices and can run on any hardware [2][3]. Product Overview - AutoGLM 2.0 is powered by domestic models GLM-4.5 and GLM-4.5V, showcasing comprehensive capabilities in reasoning, coding, and multimodal tasks [2][9]. - The product allows users to perform various tasks across multiple high-frequency applications like Meituan, JD.com, and Douyin, effectively transforming AI from a chat tool to a fully functional assistant [6][7]. - It operates independently in the cloud, enabling users to engage with other applications simultaneously, thus enhancing productivity [6][7]. Market Position - AutoGLM is positioned as a leading product in the domestic market, with expectations to drive significant growth for Zhipu [3][4]. - The annual recurring revenue (RRR) for Manus has reached $90 million, indicating a strong market presence and potential for further growth [2]. Technological Advancements - AutoGLM 2.0 represents a qualitative leap, capable of executing tasks autonomously without user intervention, adhering to the 3A principles: Around-the-clock operation, Autonomy without interference, and Affinity across devices [8]. - The product's capabilities can be integrated into various hardware through APIs, allowing for a broader application beyond just mobile and computer devices [8][9]. Performance Metrics - In benchmark tests, AutoGLM outperformed competitors like ChatGPT Agent, UI-TARS-1.5, and Claude Sonnet 4 in device use, showcasing its robustness and versatility [10].

Core Viewpoint - The article introduces the world's first universal mobile agent, AutoGLM, developed by Zhipu AI, which allows users to perform tasks on their mobile devices through voice commands, significantly enhancing convenience and intelligence [5][6][9]. Group 1: Product Features - AutoGLM operates in the cloud, enabling seamless task execution without affecting the performance of other applications on the user's device [9][33]. - The agent can handle various tasks categorized into "lifestyle assistant" and "office assistant," allowing users to interact with it as if they were using a normal smartphone [11][15]. - Users can initiate complex tasks, such as comparing prices across multiple e-commerce platforms, with minimal input required [19][20]. Group 2: Technological Advancements - AutoGLM represents a significant upgrade from traditional chatbots by executing tasks autonomously rather than merely providing instructions [31]. - The cloud execution model alleviates the burden on local devices, ensuring that users can continue using their devices without interruption [36][37]. - The integration of a cloud computer allows AutoGLM to perform high-complexity tasks that local devices may struggle with due to limited processing power [36][41]. Group 3: Industry Implications - The launch of AutoGLM aligns with a growing trend in the industry towards cloud-based agents, as seen with other major players like Alibaba Cloud [38][40]. - The product validates the feasibility and reliability of cloud execution in the agent space, potentially setting a new standard for future developments [53][54]. - AutoGLM's capabilities reflect a shift in user interaction with machines, moving from simple communication to direct task execution [55][56].

Investment Rating - The report does not explicitly state an investment rating for the industry Core Insights - The impact of AI is not limited to software; various sectors are witnessing the rise of software companies seizing opportunities in the AI era, such as Applovin in advertising and Figma and Canva in visual design [1][15] - Companies with strong know-how, proprietary data, complex processes, or regulatory barriers are less likely to be disrupted by large models; instead, these models may enhance their competitive advantages [2][20] - The development of open-source models is beneficial for software companies, allowing them to develop independently or negotiate better with closed-source models [19] Industry Trends - The report highlights a significant growth in AI-related revenues, with OpenAI's annual recurring revenue surpassing $13 billion and Anthropic's revenue reaching $4 billion, a fourfold increase since the beginning of the year [12] - Concerns about AI disrupting software have led to stock declines in companies like Adobe (down 23%) and ManpowerGroup (down 30%) [14] - The report identifies three types of AI agents: user-created agents, vendor-provided agents, and enterprise-deployed agents, indicating a shift towards personalized and automated solutions [3][37] Recommendations - The report suggests focusing on companies involved in computing power, such as Cambrian, Hygon Information, and others, as well as those developing AI agents like Alibaba and Tencent [7][53] - It also mentions companies in the autonomous driving sector, including Jianghuai Automobile and Xiaopeng Motors, as potential investment opportunities [54]

Investment Rating - The report assigns a "Buy" rating for the company for the first time [2]. Core Insights - The company is a leading AI SaaS marketing and sales platform, focusing on digital and intelligent marketing solutions for various industries, including retail, automotive, finance, healthcare, and cross-border e-commerce [8][13]. - The company aims to accelerate business growth through three strategic initiatives: building an AI-Agentforce platform, pursuing acquisitions to enhance its product ecosystem, and expanding into global markets [8]. - The report forecasts significant revenue growth, with expected revenues of 2.355 billion RMB in 2025, 3.085 billion RMB in 2026, and 4.062 billion RMB in 2027, alongside a return to profitability with net profits of 96 million RMB in 2025, 207 million RMB in 2026, and 354 million RMB in 2027 [7][8]. Summary by Sections Company Overview - Established in 2009, the company has evolved into a global leader in AI SaaS marketing solutions, with a focus on digital transformation for enterprises [8][13]. - The company has developed a comprehensive product matrix, including T Cloud for SMBs and Zhenke for large enterprises, and is enhancing its offerings with AI capabilities [17][19]. Market Performance - The company has experienced a significant revenue increase from 2.7 billion RMB in 2019 to 15.6 billion RMB in 2024, with a CAGR of 42% [19][20]. - The SaaS business is projected to account for approximately 54% of total revenue in 2024, with a recurring revenue model showing strong retention rates [17][19]. Financial Projections - Revenue is expected to grow at a compound annual growth rate (CAGR) of 51.09% from 2024 to 2025, with net profit margins improving significantly [7][19]. - The report anticipates a stable gross margin for SaaS services, close to 90%, while the precision marketing service is expected to maintain a gross margin of around 15% [17][19]. Strategic Initiatives - The company plans to leverage AI technology to enhance customer engagement and operational efficiency, with a focus on expanding its customer base from SMBs to larger enterprises [8][19]. - The report highlights the potential of the marketing and sales SaaS market in China, projected to reach 46.3 billion RMB by 2025, with significant growth opportunities in the Agent market [41][44]. Competitive Position - The company is recognized as the largest provider of marketing and sales SaaS solutions in China, holding a market share of 2.6% as of 2022 [44][45]. - The competitive landscape is characterized by a fragmented market, with the company positioned to benefit from its established customer base and technological advantages [44][45].

Core Viewpoint - The AI era is accelerating the evolution of the domestic software industry, with Kingdee International reporting a revenue increase and narrowing losses, indicating a positive outlook for profitability driven by cloud subscription and AI efficiency improvements [2][3]. Financial Performance - Kingdee International achieved a revenue of approximately 3.192 billion RMB, a year-on-year increase of about 11.2% [2]. - The company reported a loss attributable to equity holders of approximately 98 million RMB, a year-on-year reduction of about 55.1% [2]. - Cloud subscription revenue reached approximately 1.684 billion RMB, growing by about 22.1% year-on-year [3]. - The gross margin improved by 2.4 percentage points to approximately 65.6%, primarily due to the increased proportion of cloud subscription revenue [3]. Business Model Evolution - Kingdee has transitioned to a subscription-based cloud service model, with annual recurring revenue (ARR) from cloud subscriptions growing by 18.5% to 3.73 billion RMB [3]. - The company aims to increase the proportion of cloud subscription business, delegating implementation tasks to third-party partners to enhance efficiency and flexibility [3]. - The gross margin for cloud subscriptions reached 96.2%, with expectations for further improvement as IT technology advances [3]. AI and Product Development - Kingdee launched several AI-native products, including the Cangqiong AI Agent platform 2.0, and is focusing on multi-agent collaboration for large enterprises [4][5]. - The company is addressing the high internal management requirements of large enterprises by offering customizable AI solutions [5]. - Kingdee emphasizes the importance of data security and reliable business data management in advancing AI applications [7]. Market Trends and Challenges - The year is recognized as the "Agent Year," with various vendors entering the market, but challenges such as renewal willingness and potential product homogenization are emerging [6]. - Kingdee is strategically selecting AI products based on high-frequency demand and collaboration with enterprises for co-creation [7]. - The company identifies data quality, technical talent shortages, and ecosystem development as key challenges in implementing AI solutions [8]. International Expansion - Kingdee is actively investing in overseas markets, particularly in Southeast Asia and the Middle East, with plans for product localization [9]. - The company has already released product packages for 14 countries and regions, with potential expansion into Africa, Europe, and Japan in the future [9].

而对于万千具体场景中的企业而言，Agent「自主执行并管理各类任务」的角色转变，意味着一场新的生产力变革，没有人想被时代落下，于是都开始轰轰烈烈构 建起属于自己的 Agent。 然而，事情没那么简单。很多企业部署了 Agent 之后，发现并没有达到预期效果，现实与理想之间的巨大落差开始让他们疑惑：难道 Agent 只是一场夸大的「纸上 谈兵」？ 无疑，技术的进步肉眼可见，Agent 的实用也并非虚假宣传，这是出现这种情况更深层的原因在于，业界对 Agent 平台的狂热追捧下陷入一个误区：把 Agent 平 台、Bot 框架等当作 AI Infra。 机器之心报道 机器之心编辑部 「新的风暴已经出现！」 当我们谈论 AI Infra 的时候，我们在谈论什么？ 年初，DeepSeek 前脚带来模型在推理能力上的大幅提升，Manus 后脚就在全球范围内描绘了一幅通用 Agent 的蓝图。新的范本里，Agent 不再止步于答疑解惑的 「镶边」角色，开始变得主动，拆解分析需求、调用工具、执行任务，最终解决问题…… 这质的变化引起的效应如投石入水，激起的涟漪不断向外蔓延……Agent 成为 2025 年 AI 的主流叙事 ...