证券研究报告 | 浙商早知道 市场总览 重要观点 报告日期：2025 年 06 月 19 日 浙商早知道 2025 年 06 月 19 日 ：王禾 执业证书编号：S1230512110001 ：021-80105901 ：wanghe@stocke.com.cn http://www.stocke.com.cn 1/4 请务必阅读正文之后的免责条款部分 ❑ 大势：6 月 18 日上证指数上涨 0.04%，沪深 300 上涨 0.12%，科创 50 上涨 0.53%，中证 1000 下跌 0.1%，创业板 指上涨 0.23%，恒生指数下跌 1.12%。 ❑ 行业：6 月 18 日表现最好的行业分别是电子（+1.5%)、通信（+1.39%)、国防军工（+0.95%)、银行（+0.92%)、 电力设备（+0.32%)，表现最差的行业分别是美容护理（-1.73%)、房地产（-1.35%)、建筑材料（-1.22%)、非银金 融（-1.16%)、轻工制造（-1.13%)。 ❑ 资金：6 月 18 日全 A 总成交额为 12217.64 亿元，南下资金净流入 12.42 亿港元。 ❑ 【浙商煤炭 樊金璐】煤炭 半年行业策略 ...

Hey everybody, my name is Eno, co-founder and CTO of a company called Factory. Uh, at Factory, we believe that the way we build software is radically changing. We are transitioning from the era of human-driven software development to agent-driven software development.You can see glimpses of that today. However, it seems like we are trying to get to that future incrementally. Uh the current zeitgeist is to take uh the IDE, a tool that was designed first and foremost for human beings to write lines of code by ...

21世纪经济报道记者王峰 北京报道 AI报志愿究竟靠不靠谱？ 高考志愿填报在即，"天价报志愿"服务再次受到欢迎。据报道，网红张雪峰旗下机构两款12999元和 18999元的志愿服务产品早早售罄。 "天价报志愿"服务只能满足极少数考生的需求，在提供普惠性、基础性志愿服务方面，AI曾被寄予厚 望，但此类产品面世几年来，要么错误较多，不同产品推荐结果相互打架，要么只能作为参考，考生依 然需要志愿规划师的指导。 2025年或将有所改变。深度思考技术推动大模型辅助志愿填报又进了一步，不仅所推荐的志愿准确率更 高，而且高考志愿大模型有了AI Agent的雏形，搭建了类似真人志愿规划师的工作流，强化了志愿填报 的规划性。 普惠的AI技术越发展，高考志愿服务市场就越理性，高考考生越能远离"天价报志愿"。 不过，AI高考志愿还无法完全取代真人志愿规划服务，推动高考志愿填报服务的普及、普惠，需要加 大公共服务力度。 AI高考志愿进阶之路 大模型如何改变AI高考志愿产品？ 2024年以前，市场上的AI高考志愿产品还不是大模型技术，而是基于数据库筛选的大数据技术。 考生输入自己的地区、选科、分数、排名信息，以及意向高校和专业的所在地 ...

Advanced AI agents from Agora and WIZ.AI can power call centers with multilingual support and contextual understandingSANTA CLARA, Calif., June 18, 2025 /PRNewswire/ -- Agora (NASDAQ: API), a global leader in real-time engagement and conversational AI technology, today announced its partnership with WIZ.AI to bring enterprise-grade AI agent solutions to the market.WIZ.AI, a leading provider of AI agents, is partnering with Agora to power enterprise-grade AI agents that offer local language support and conte ...

Core Insights - MiniMax has launched its first open-source inference model, M1, which, despite average benchmark performance, boasts the industry's longest context capabilities with 1 million tokens input and 80,000 tokens output [2][52]. - The company aims to regain its competitive edge in the AI sector, particularly with the anticipated rise of agents in 2025 [4][70]. - M1's strengths lie in its long context window and reasoning capabilities, making it suitable for agent applications, although its overall performance remains average compared to leading models [30][29]. Group 1: Model Capabilities - M1's inference model exhibits a long reasoning chain, similar to other recent domestic open-source models, but this can lead to output inaccuracies [6]. - The model successfully translated a 33-page PDF while maintaining formatting, showcasing its long context capabilities [22][23]. - M1's performance in coding tasks is on par with top-tier models, indicating it has entered the first tier of open-source models [21]. Group 2: Agent Development - MiniMax is currently testing its general-purpose agent, which shows improved front-end performance and project delivery [31][32]. - The agent can gather information through extensive web searches and validate its outputs by testing the developed websites [37][39]. - The agent's ability to utilize browser tools for self-assessment is a notable innovation compared to traditional agents [36]. Group 3: Technical Architecture - M1 features a hybrid architecture centered on a lightning attention mechanism and an efficient reinforcement learning algorithm called CISPO [51][57]. - The model's training efficiency is remarkable, requiring only 512 H800 chips and three weeks, costing approximately $534,700, significantly lower than typical large model training costs [63][64]. - M1's input and output capabilities provide a competitive edge in long-context applications, particularly for agent functionalities [66][68]. Group 4: Market Position and Future Outlook - The trend towards agent development in 2025 presents an opportunity for MiniMax to leverage its long-context model [70][72]. - The success of agents will depend on various factors, including end-to-end capabilities, tool utilization, and the performance of the primary model [75][78]. - MiniMax's technological advantages in long context processing position it favorably in the competitive landscape, but the ultimate success will hinge on translating these advantages into user value [78].

前两个月我介绍了几个给 AI Agent 的专属浏览器产品，其中 Browserbase 增长尤为快速，仅过去一年就 完成了 3 轮融资《 给 AI Agent 的专属浏览器已 3 亿美金估值，8 位华人团队创意 AI 1200 万美金 ARR 正 融资 》。 当时我在文章里说 Browserbase 已经再次以 3 亿美金估值完成了 B 轮的融资，由 Notable Capital 领投。 今天，Browserbase 正式官宣了此次融资，估值 3 亿美金，领投方正是 Notable Capital，而金额为 4000 万美金。 今天，OpenAI CEO Sam Altman 说，Meta 为了挖 OpenAI 的人才，直接开出了高达 1 亿美金的薪酬。因 此作为 Meta 的竞争对手，估计 OpenAI、Anthropic 等可能都会考虑后续与 Scale AI 的合作关系。 这就给新兴玩家带来巨大机会，除了像 Mercor 这种新兴的 AI 招聘平台外，前两天刚介绍的这个传统招聘 平台《 AI 让传统招聘平台年增 1 亿美金 ARR，Glean 估值 72 亿美金了 》以及另外几个都声称需求爆 增。 ...

AI 正在深刻重塑软件研发的全过程，从代码补全到流程协同，从"写工具"到"用工具"的范式转变已在多个技术团队中开始落地。 为帮助更多研发团队理解和借鉴这些转变， 6 月 27～28 日 ，在 AICon 全球人工智能开发与应用大会·北京站 上，多位来自一线大厂的技术大牛将带 来关于"AI 赋能研发提效"的硬核分享，呈现他们在 智能研发、AI 驱动协作与开发效能提升 等方面的思考与经验。 以下为重点演讲前瞻，一起提前了解他们将带来的分享内容： 《从 Copilot 到 Agent：AI 编程的范式革新》 by 丁宇（叔同）｜阿里云智能集团 研发副总裁 《万人研发规模代码智能化落地实践及发展思考》 by 张乐｜腾讯 资深技术专家 腾讯在万人研发团队中推动代码智能化的过程中，积累了大量一线经验。本次分享将聚焦代码补全、技术对话、代码评审、单元测试等多个维度的 智能化落地路径与技术架构演进。张乐将对行业内外的不同路径进行横向对比，剖析目前哪些方向已经取得实质进展、哪些仍处于探索期，并结合 自身实践提供判断落地可行性的标准和技术选型思路。 《Coding Agent 驱动研发提效：从实践到流程改进》 by 杨方伟｜同程 ...

Group 1 - The core viewpoint is that the gaming industry is expected to see accelerated application and commercialization of AI products, particularly focusing on AI Agents, AI companionship, and AI multimodal technologies [1] - AI Agents are viewed as productivity tools that enhance efficiency through autonomous decision-making and dynamic interaction, with expectations for continuous optimization throughout the year [1] - AI companionship addresses personalized interaction needs and falls within the broader entertainment sector [1] Group 2 - AI multimodal technologies, including audio, video, and 3D models, are undergoing continuous iteration, driving the accelerated implementation of industry applications [1] - The gaming ETF (code: 516010) tracks the animation and gaming index (code: 930901), which is compiled by China Securities Index Co., Ltd., reflecting the overall performance of listed companies in the Chinese animation and gaming industry [1] - The index constituents are primarily distributed across cultural media and software development sectors, showcasing both industry concentration and innovative growth characteristics [1]

海外科技厂商 AI 布局与 To B Agent 进展 20250617 摘要 微软 Azure 云服务凭借强大的 GPU 算力及 AI Foundry 平台对多种开源 模型的支持，在 AI 基础设施建设方面具有显著优势，尤其在 ToB 场景 和边缘计算领域表现突出，构建了全面的 AI 开发工具链。 亚马逊主要通过 AWS 云服务推动 AI 发展，侧重于算力支持和图像模型 服务，尤其面向中小企业提供 DeepSeek、LLAMA 等模型的部署方案， 并通过 API 销售云模型。 Meta 近期推出 LLAMA4 并收购 Scale AI，旨在通过优化数据层提升模 型能力，但目前效果尚未显著，其早期在开源领域的贡献为后续发展奠 定基础。 AI Agent 开发的核心壁垒在于垂直领域专家知识和私有数据，工程化解 决方案如 Workflow 管理是关键技术，而生成式 AI 并未引入全新技术， 更多是对已有技术的优化应用。 微软 Copilot 系列产品在 M365 办公场景中应用广泛，但在整合能力和 多模态能力方面发展低于预期，Word 和 Excel 表现较好，PowerPoint 因视觉元素处理能力不足评价较低 ...

With the rise of MCP servers, A2A, and our trusty friend, OpenAPI, it turns out the web browser may be the default MCP server for the rest of the internet. In this talk, we'll walk through how a web browsing tool is probably the only tool you'll need to enable production AI Agents. About Paul Klein IV Paul Klein IV is a San‑Francisco‑based serial entrepreneur and engineer. After honing his chops at Twilio during it's IPO and founding Stream Club—a live‑streaming platform acquired by Mux in 2021 he launched ...