云知声(09678)再涨近19%，近四个交易日累涨近90%。截至发稿，涨16.75%，报416.8港元，成交额4.69 亿港元。 消息面上，云知声此前发布公告称，公司预期于2025年内实现大模型相关业务收入约6.0亿元至6.2亿 元，大模型相关业务收入相较2024年同比增长约1057%至1095%。此番业绩增长的原因是大模型核心技 术能力领先、商业化落地进程加速。 值得注意的是，近日，云知声正式对外发布"山海.知音"大模型2.0版本，标志着企业在"一基两翼"技术 战略布局上完成了关键拼图。依托"山海.Atlas"智算基座，通过端到端架构重塑语音交互逻辑，标志着 AI语音交互正从"听懂指令"迈向"共情对话"的新阶段。 ...

Core Insights - Cloud Wisdom (云知声) has successfully transitioned from "technological leadership" to "product implementation" and "commercial realization," marking a significant achievement in the AI industry [1] - The company's revenue from large model-related businesses is projected to reach between 600 million to 620 million RMB in 2025, representing a year-on-year increase of 1057% to 1095%, and is expected to account for 48% to 53% of total revenue [1] Group 1: Growth Drivers - The company's growth is attributed to three main drivers: technological leadership, enhanced productization capabilities, and deep integration into serious application scenarios [2] - Cloud Wisdom has developed a comprehensive "Shan Hai" large model matrix, which includes general large language models, multimodal models, and vertical industry models, achieving top rankings in various assessments [2] - The newly released "Shan Hai·Zhi Yin 2.0" has reached the highest level in speech recognition, supporting over 30 Chinese dialects and 14 international languages, while also achieving advanced speech synthesis capabilities [2] Group 2: Focus on High-Value Applications - The company focuses on high-value "serious scenarios" such as healthcare, insurance, and transportation, which require high accuracy, compliance, and stability [3] - The "Shan Hai·Zhi Yi" system has been deployed in nearly 400 hospitals, with a high case generation citation rate of nearly 90% at Beijing Friendship Hospital, indicating strong commercial value [3] Group 3: Strategic Direction - Cloud Wisdom aims to become a leader in the professional large model and intelligent agent field in China by 2026, with a clear strategic direction [4] - The company plans to deepen its "one base, two wings" technology strategy, focusing on solidifying its general large model foundation and developing expert-level intelligent agents [4] - The successful model in the healthcare sector will be replicated in other complex fields such as insurance, government, and transportation, aiming for scalable commercial potential [4] Group 4: Industry Implications - Cloud Wisdom's transformation reflects the broader trends in the Chinese AI industry, emphasizing that market value is derived from solving real problems rather than just parameter scale [5] - The case of Cloud Wisdom provides key insights for investors, highlighting the importance of companies with controllable technology, clear application scenarios, and visible revenue [6]

Core Insights - The company is accelerating its "one foundation, two wings" technology strategy amid the rise of intelligent agents, recently launching the "Shan Hai · Zhi Yin" model 2.0 after upgrading the "Shan Hai · Zhi Yi" medical model 5.0 [1] Group 1: Model Capabilities - The "Shan Hai · Zhi Yin" model 2.0 focuses on three major capability evolutions: understanding professional and local dialects, expressing warmth and emotional connection, and achieving rapid responsiveness [1] - In terms of "understanding," the model's ASR (Automatic Speech Recognition) capabilities have demonstrated leading performance in both public test sets and proprietary full-scene test sets, surpassing mainstream domestic open-source and closed-source speech models, reaching the highest industry standards [1] - For the "expression" aspect, the Shan Hai · Zhi Yin-TTS (Text-to-Speech) features a "highly human-like and creatively diverse" core, currently supporting 12 dialects (including Cantonese, Sichuanese, and Shanghainese) and 10 foreign languages, with the ability to switch between 12 styles of Mandarin [1] Group 2: Technological Foundation - The capabilities are underpinned by the company's proprietary "Shan Hai · Atlas" intelligent computing foundation, which deeply integrates a general multimodal model base with the Atlas architecture, serving as the foundation for professional intelligent agents and the core of perceptual AI [2]

在"表达"层面，山海.知音-TTS以"高度拟人+创意多元"为核心，目前支持12种方言(粤语、四川话、上 海话全拿下)+10种外语，甚至能切换12种普通话风格。 更重要的是，山海.知音2.0基于端到端交互大脑，攻克了流畅全双工交互的挑战，支持随时打断、即时 接话与连贯追问，使人机对话如挚友交谈般行云流水。 这一切能力的背后，是云知声独创的"山海.Atlas"智算一体基座，将通用多模态大模型底座与Atlas基础 架构深度整合，既是专业智能体的基础，也是感知AI中枢的根基。 伴随智能体时代浪潮奔涌，云知声(09678)正加速完成其"一基两翼"技术战略的能力拼图。继年前升 级"山海.知医"5.0医疗大模型后，公司于近日重磅发布"山海.知音"大模型2.0。 据悉，"山海.知音"大模型2.0依托"山海.Atlas"的多模态、跨语言基座能力，核心聚焦三大能力进化—— 听懂专业与乡音、聊出亲情与温度、极致机敏反应。 在"听懂"的维度，此次模型ASR能力在公开测试集和自有全场景测试集中，均显现了领先的语音识别能 力，评测中实现了从通用到极端全面的领先水平，超过了国内主流的开源和闭源语音大模型，达到业界 最高水平。 ...