当大模型的聚光灯照向实体经济，工地上挥汗如雨的基建工人、在营业厅忙碌的金融从业者、在流水线 上日夜赶工的制造业工人，都可能面临同一个问题：论文里惊艳的大模型技术，究竟怎样才能变成车 间、工地、营业厅里实实在在的生产力? 从参数狂欢、开源大战到价值突围 前两年，科技巨头们还在参数规模的赛道上竞速，千亿级甚至万亿级模型接连登场，论文里的数字不断 刷新人类对"智能极限"的想象;如今，企业主们却陷入更深一层的思考 ——部分企业引入大模型后，却 未能实现预期的业务增长。 2025年，参数比拼不再是行业焦点，开源生态的争夺取而代之。头部企业纷纷将核心模型代码、数据开 放共享，试图通过开发者社区的力量，加速技术迭代与场景落地。这场从"闭门造车"到"开源共建"的转 变，标志着大模型行业正从追求技术，迈向挖掘商业价值的新阶段。 头部通用大模型企业开始加速向垂直行业渗透。华为云深入钢铁冶炼场景，通过对钢材成分、工艺进行 大数据分析，以及在线监控、实时调整优化工艺，将产品性能合格率提升。科大讯飞打造医疗大模型， 基于患者症状、检查数据等，快速生成鉴别诊断建议，极大提升基层医疗服务能力 。阿里云的通义大 模型已经下矿，矿山重大风险辨 ...

Core Viewpoint - Zhongguancun KJ has made significant progress in the Xinchuang field by successfully obtaining Huawei Ascend certification for its self-developed Dazhu model application development platform, marking a comprehensive adaptation to the Huawei Ascend ecosystem and enhancing its capabilities in various sectors such as finance, government, and enterprise services [1][2]. Group 1 - The Dazhu model platform integrates Huawei Ascend AI processor capabilities, forming a complete development capability that includes computing factories, data factories, model factories, and intelligent agent factories [2]. - The platform supports flexible access to high-performance AI computing resources from Huawei Ascend Cloud and is compatible with mainstream domestic chips and hardware, creating a full-stack domestic closed loop from chip to application [2]. - Zhongguancun KJ has launched an all-in-one Dazhu model application development machine to simplify deployment processes for enterprises, featuring pre-installed frameworks for domestic Xinchuang environments and supporting rapid tuning of mainstream open-source models [2]. Group 2 - Zhongguancun KJ and Huawei Cloud jointly released the "Ascend Cloud + Dazhu Model Platform" solution, creating an integrated service system that combines industry knowledge, business processes, and general large models to quickly train high-precision vertical models [3]. - The company has developed the first large model application platform in the engineering sector, 'Lingzhu', in collaboration with Ningxia Jiaojian Transportation Technology Research Institute, supporting various innovative scenarios [5]. - In the financial sector, Zhongguancun KJ partnered with CITIC Construction Investment Securities to create a "large model intelligent training platform," enhancing compliance learning services for over 300 branch client managers [5]. Group 3 - Zhongguancun KJ is advancing the usability of domestic large models through self-research, application co-creation, and ecosystem building, aiming to accelerate the implementation of vertical large model applications [6]. - The company plans to continue collaborating with ecosystem partners to develop more scenario-based solutions, supporting the prosperity of the domestic Xinchuang ecosystem and the intelligent upgrade of industries [6].

Group 1 - The event "AI+ Innovation Cooperation Meeting" was held in Nanning, Guangxi, focusing on AI collaboration between Beijing and Guangxi [1][5] - Chen Gang, Secretary of the Guangxi Zhuang Autonomous Region, emphasized the importance of AI capability construction and encouraged enterprises to participate in the China-ASEAN AI Innovation Cooperation Center [3][6] - Zhongguancun KJ has actively engaged in Guangxi's "AI+" initiatives, conducting industry research and launching various AI model projects tailored to local industries [2][4] Group 2 - Zhongguancun KJ is recognized as the only company leveraging domain-specific models to be listed in the "2024 Hurun China AI Companies Top 50" [4] - The company employs a "platform + application + service" strategy to accelerate the implementation of vertical AI models across various sectors [4] - Notable collaborations include the development of the "Lingzhu" platform for the engineering sector and the "Baihe" model for the shipbuilding industry, showcasing comprehensive AI solutions [4]

编辑部 发自 凹非寺 量子位 | 公众号 QbitAI 大模型技术加速向产业渗透，如何直击业务痛点、带来真实增效？ "平台＋应用＋服务"是企业大模型落地的最佳路径。 在第三届AIGC产业大会上， 中关村科金总裁喻友平 分享如上方法论。 即使看似简单的需求，也需要经历需求拆解、数据调优与流程重构的闭环。在这个过程中，企服厂商需要提供好服务。 为了完整体现喻友平的思考，在不改变原意的基础上，量子位对演讲内容进行了编辑整理，希望能给你带来更多启发。 中国AIGC产业峰会是由量子位主办的AI领域前沿峰会，20余位产业代表与会讨论。线下参会观众超千人，线上直播观众320万+，累计曝光 2000万+。 话题要点 认知型AI亦遵循"C端先行、B端深化" 大模型技术正从消费端向产业端加速渗透； 认知性AI同样遵循"C端先行、B端深化"的路径； 企业需要端到端的解决方案而非孤立技术模块； 算法开源趋势使得数据主权愈加重要； 企业大模型落地最佳路径就是做好"平台+应用+服务"。 各位嘉宾下午好，非常荣幸受邀参与量子位大会的分享。此前各位专家已就前沿技术展开深度探讨，我的主题则聚焦于 大模型在企业服务领 域的落地实践——如何通过技 ...