以下文章来源于投中网 ，作者张雪 投中网 . 投中网是领先的创新经济信息服务平台，拥有立体化传播矩阵，为创新经济人群提供深入、独到的智识 和洞见，在私募股权投资行业和创新商业领域拥有权威影响力。官网：www.chinaventure.com.cn 本文来自微信公众号： 投中网 （ID：China-Venture） ，作者：张雪，题图来自：AI生成 市场上似乎没有哪家AI公司像Manus这样，从一面世就在盛誉与争议间摇摆。尽管外界评价对这家 AI Agent公司相当分化，但自3月初仅限邀请用户上线以来，Manus因能够处理复杂任务的惊艳表 现，吸引了全球人工智能界的关注，这使得它成为继DeepSeek之后，最热门的中国人工智能初创企 业。 今年4月，据彭博社报道，Manus完成了由Benchmark领投的新一轮融资，金额为7500万美元 （约 5.4亿人民币）， 估值也达到了5亿美元 （约36亿人民币） 。后来虽然经历波折，据我所知，这笔融 资已经顺利到账。 近日，据一位正在帮Manus融资的投资人，Manus已经接近完成新一轮人民币融资，由某地国资领 投，金额达数亿元，投前估值37亿人民币。投中网向Manus ...

进入 2025 年， AI 已经从一个前沿概念，变为驱动生产力变化的核心工具之一 ， 并以 一种前所未有 的方式，影响着全球的商业社会。从企业的战略规划到员工的日常工作 。 而 中国的人工智能产业呈现出一种特别的景象：一方面， 虽然出现了 Deepseek 这样的"黑马"模型， 但 在以大语言模型（ LLM ）为代表的基础 AI 技术研发上，中国企业 仍然 面临着技术追赶、成本高 昂、人才稀缺的挑战，可 谓 " 炼 AI 难 " 。 但当我们将目光转向 AI Agent （人工智能代理）时，会发现凭借独特的市场环境和产业基因，中国正 迎来一片充满机遇的领域 。 甚至 可以说 在中国， 做 Agent 是一条更为容易的坦途 。 什么是AI Agent？ 从"数字助理"到"数字员工"的进化 要理解 AI Agent ，我们可以从一个具体的应用场景说起 。 想象一下： 你 是一家电商公司的运营经理，明天上午 10 点需要针对上周销量意外下滑的爆款商品 " xx 智能降噪耳机 " 开一个紧急复盘会。 你 需要手动登录公司的 ERP 和 BI 系统，导出销售数据、用户画像数据和库存周转数据；然后，打开 天猫、京东等竞品 ...

近日，据一位正在帮Manus融资的投资人，Manus已经接近完成新一轮人民币融资，由某地国资领投，金额达数亿元，投前估值37亿人民币。 投中网向 Manus高层确认此消息，得到回复"这个完全是谣言"。 不过，这位据称正在帮Manus融资的投资人向我表示，由于签署了保密协议，目前不方便公开更 多细节，但"保证一定会投"，还向我提供了两份BP。 一级市场虚虚实实，融资消息真实性我将持续关注。两份 Manus 的 BP （融资计划书），分别对应美元和人民币机构。 先来大概述说 Manus 的近况。 半年估值就翻5倍。 作者丨 张雪 来源丨 投中网 市场上似乎没有哪家 AI 公司像 Manus 这样，从一面世就在盛誉与争议间摇摆。尽管外界评价对这家 AI Agent 公司相当分化，但自 3 月初仅限邀请 用户上线以来，M anus 因能够处理复杂任务的惊艳表现，吸引了全球人工智能界的关注，这使得它成为继 DeepSeek 之后，最热门的中国人工智能 初创企业。 今年 4 月，据彭博社报道， Manus 完成了由 Benchmark 领投的新一轮融资，金额为 7500 万美元（约 5.4 亿人民币），估值也达到了 5 ...

举个例子，针对北京市高考海淀区模拟全卷测评，豆包1.6相对去年版本的表现，理科成绩显著提升了 154分，文科提升了90分。 这也就是客户都喜欢用豆包的原因，有数据侧面印证，今年3月份，豆包大模型的日均调用数是12.7万 亿tokens。而截至今年5月底，这个数字已经超过了16.4万亿tokens。 根据IDC的数据显示，火山引擎在中国公有云大模型服务调用量上也稳居榜首，市场份额46.4%，接近 一半。 在市场规模稳居第一的同时，豆包大模型1.6延续了以前的价格优势。通过技术和商业的双重创新，豆 包1.6首创按"输入长度"区间定价，深度思考、多模态能力与基础语言模型统一价格。 在企业使用量最大的0-32K输入区间，豆包1.6的输入价格为0.8元/百万tokens、输出8元/百万tokens，综 合成本只有豆包1.5深度思考模型或DeepSeek R1的三分之一。Seedance1.0 pro模型每千 tokens仅0.015 元，每生成一条5秒的1080P视频只需3.67元。 加速Agent应用落地，重塑行业用户体验 近日，豆包大模型发布全新的1.6模型，其综合能力"火力全开"。新版本不仅在推理、数学、指令遵循 ...

Group 1 - The article highlights the importance of weekly announcements from Sunday to Thursday, which include significant stock market updates such as suspensions, increases or decreases in holdings, investment wins, acquisitions, earnings reports, unlocks, and high transfers [1] - It emphasizes the collaboration of a company with Shanghai Museum and other IPs to develop a series of cultural and creative products, as well as the establishment of a joint venture primarily focused on producing military silencing materials [1] - Another company is noted for integrating anti-quantum password algorithms and quantum random number chips into its products, showcasing advancements in military, quantum technology, cloud computing, digital currency, blockchain, AI, and chips [1] Group 2 - A company is investing nearly 6 billion in the optical communication sector, indicating a strong commitment to robotics, new energy vehicles, and optical modules [1]

Quantitative Models and Construction Methods - **Model Name**: Report/Document Interpretation Workflow - **Model Construction Idea**: Automate the process of interpreting financial reports and extracting key information, including formulas, using AI agents and workflows[28][30] - **Model Construction Process**: 1. Use Coze's official file-reading plugin to extract document content and formula structures[30] 2. Configure prompt logic and output format using LLM nodes in the workflow[30] 3. Test the workflow by inputting a URL of a quantitative research paper, where the AI agent summarizes key information and accurately interprets formulas[31] - **Model Evaluation**: Demonstrates the ability to process complex financial documents and provide accurate formula interpretations, enhancing efficiency in financial research[31] - **Model Name**: Real-Time Financial Data Analysis Workflow - **Model Construction Idea**: Automate the retrieval and analysis of real-time financial data from web sources or plugins[35][38] - **Model Construction Process**: 1. Construct a workflow with a code-processing node to generate complete URLs based on user-input stock codes[38] 2. Use a data-scraping node to retrieve real-time financial data from websites like Sina Finance[35][38] 3. Input the data into the DeepSeek LLM node for comprehensive analysis, focusing on profitability, solvency, and operational efficiency[39] - **Model Evaluation**: Provides timely and structured financial insights, enabling informed decision-making in investment analysis[39] - **Model Name**: Research Report Summarization Workflow - **Model Construction Idea**: Automate the process of extracting and summarizing content from multiple research reports or news articles[52][55] - **Model Construction Process**: 1. Use Coze plugins to scrape HTML content from websites like Eastmoney[55] 2. Employ loop nodes to process multiple reports and extract relevant content[55] 3. Store the extracted data (e.g., titles, content, institution names, links) in Feishu multi-dimensional tables for further analysis[57] - **Model Evaluation**: Effectively consolidates and organizes large volumes of research data, improving accessibility and usability for financial analysts[57] Model Backtesting Results - **Report/Document Interpretation Workflow**: Successfully summarized key information and accurately interpreted formulas from a quantitative research paper[31] - **Real-Time Financial Data Analysis Workflow**: Generated detailed financial analyses based on real-time data, covering multiple financial metrics such as ROE, net profit, and cash flow[39][48] - **Research Report Summarization Workflow**: Efficiently extracted and stored structured data from multiple research reports, enabling streamlined analysis and reporting[57][60] Quantitative Factors and Construction Methods - **Factor Name**: None explicitly mentioned in the report Factor Backtesting Results - **Factor Results**: None explicitly mentioned in the report

Core Insights - Quark has launched China's first AI model specifically designed for college entrance examination (Gaokao) application scenarios, featuring three core functions: deep search, application report, and intelligent application selection [3] - The model was fine-tuned by benchmarking against industry experts, ensuring that it meets the unique requirements of Gaokao application, which demands high accuracy and logical coherence [3][4] - The fine-tuning process involved creating a structured knowledge base covering over 2,900 universities and nearly 1,600 undergraduate programs, with a focus on data verification and cross-referencing [4] Model Fine-Tuning - The fine-tuning of the model is critical, involving targeted instruction adjustments and the collaboration of hundreds of experts to create a unique generation mechanism [4] - The team distilled thousands of past decisions from human experts to develop a reasoning chain that informs the model's decision-making process [4] - The model's design includes a mechanism to prevent "hallucinations," ensuring that the final recommendations are based on real data and historical validation [4] Agent Product Development - Quark's team served over 30 million users, with 50% being students from third-tier cities or below, indicating a significant market reach [6] - The introduction of the Agent format represents a shift in the industry, with a focus on providing expert-level advice rather than just public information [6] - The year is seen as a pivotal moment for Agent products, with major tech companies rapidly advancing their offerings in this space [6][7] Market Trends - The global AIGC technology penetration is projected to exceed 40% by 2025, with the AI Agent market expected to grow from $5.1 billion in 2024 to $47.1 billion by 2030, reflecting a compound annual growth rate of 44.8% [7]

Agent Adoption & Scalability - LinkedIn aims to scale agentic adoption within the organization to enable broader idea generation [2] - LinkedIn built the Hiring Assistant, its first production agent, to automate recruiter tasks and free up time for candidate interaction [3] - The Hiring Assistant follows an ambient agent pattern, operating in the background and notifying recruiters upon completion [4][5] - LinkedIn adopted a supervisor multi-agent architecture, with a supervisor agent coordinating sub-agents that interact with LinkedIn services [6] Technology Stack & Framework - LinkedIn standardized on Python for GenAI development, moving away from its traditional Java-centric approach [7][8] - The company built a service framework using Python, gRPC, Langchain, and Langraph to streamline the creation of production-ready Python services [9][19] - Over 20 teams have used this framework to create over 30 services supporting Generative AI product experiences [9][10] - Langchain and Langraph were chosen for their ease of use and sensible interfaces, enabling rapid development and integration with internal infrastructure [22][23] Infrastructure & Architecture - LinkedIn invested in a distributed architecture to support agentic communication modes [10] - The company modeled long-running asynchronous flows as a messaging problem, leveraging its existing messaging service for agent-to-agent and user-to-agent communication [26][27] - LinkedIn developed agentic memory with scoped and layered memory types (working, long-term, collective) [29][30] - LinkedIn implemented a centralized skill registry, allowing agents to discover and access skills developed by different teams [34][35]

Group 1 - The global industrial chain, monetary system, and asset analysis framework are undergoing reconstruction due to diminishing trust among countries, with gold potentially entering a long-term bull market driven by de-dollarization and ongoing central bank purchases [3] - Domestic economic demand remains to be boosted, and policies are expected to maintain a gradually positive tone [3] - Inflation is hovering at low levels, with the key to its rebound lying internally rather than externally, suggesting that policy efforts may become more aggressive in the second half of the year [5] Group 2 - May export growth has slowed, not due to previous over-shipments or temporary fluctuations, but rather due to the peak and subsequent decline of tariff expectations, indicating a resilient export sector despite a lower central tendency [9] - The high-interest rate environment caused by recent dollar credit discounts has led to a notable slowdown in private credit expansion in the U.S., creating a fragile balance that requires careful policy management to avoid potential debt crises [11] - The market for human-robot bearings is expected to see significant growth due to the development of humanoid robots, with domestic replacement opportunities becoming increasingly prominent [27] Group 3 - The recent trading heat in Chinese assets has increased, with a notable inflow of financing funds and new equity fund issuances exceeding 10 billion [13] - The Hong Kong stock market is emerging as a key battleground in the current bull market, driven by the scarcity of attractive assets and supportive domestic policies [16] - The expansion of ETFs is beneficial for credit bonds, with significant differences in duration and component concentration between Shanghai and Shenzhen market indices [20]

Core Insights - The report maintains an "outperform" rating for the computer industry, emphasizing the rapid iteration of AI technology and the promising outlook for AI Agents and computing power leasing [2][3][7] - The global AI Agent market is projected to reach $790 million by 2025 and grow to $52.6 billion by 2030, with a compound annual growth rate (CAGR) of approximately 46% [4] Market Trends - Google has reestablished its position in the AI sector with the launch of Gemini 2.5 Pro, which has significantly increased its token processing capacity from 97 trillion to 480 trillion tokens monthly [5][14] - Alibaba's Qwen3 model has achieved a substantial performance improvement, becoming the strongest open-source model globally, with a data scale nearly double that of its predecessor [6][26] Company Developments - Google has introduced the A2A protocol to facilitate collaboration between AI Agents, aiming to create a new ecosystem for AI applications [19][20] - Alibaba's Qwen3 model supports multiple thinking modes and has enhanced capabilities for Agent development, significantly improving its performance in various applications [32][34] Investment Recommendations - The report suggests focusing on companies actively involved in AI applications and Agent development, such as Kingsoft Office, Hehe Information, and Yonyou Network [7][8] - The computing power leasing sector is highlighted as a key area for investment, with companies like Zhiwei Intelligent and others benefiting from increased demand [7][8]