高盛发研报称，预期快手可灵AI的新模型将推动快手用户参与度持续提升，现时维持快手2025财年AI 相关收入达到1.54亿美元的预测(相对公司指引为1.25亿美元)，并重申"买入"评级，基于2026年底每股 盈利预测的约14倍，给予目标价83港元。高盛预期，可灵新模型能以更具竞争力的成本提供顶级性能的 AI影片生成方案，认为目前AI影片生成行业仍处于早期阶段，市场规模正快速增长，并可能颠覆传统 广告及短剧制作模式，而关键正在于持续升级的模型能力与用户渗透率。 来源：智通财经网 消息面上，9月23日，可灵AI基座模型再升级，视频生成推出可灵2.5 Turbo 模型，效果进一步提升。文 本理解与响应显著提升，能更深入理解具有因果关系的复杂指令，实现对创意表达的精细化控制。动态 效果进一步增强，能生成动态幅度更大的动作和运镜，在提升运动呈现幅度和强度的同时，更流畅，更 稳定。同时，可灵2.5 Turbo 模型高品质模式(1080p)下生成5秒长度的视频仅需25灵感值，相比2.1模型 同档位便宜近30%。 快手-W(01024)涨近4%，截至发稿，涨3.78%，报76.85港元，成交额12.26亿港元。 ...

Core Viewpoint - Kuaishou-W (01024) shares rose nearly 4%, with a current increase of 3.78%, trading at HKD 76.85, with a transaction volume of HKD 1.226 billion [1] Group 1: Product Development - On September 23, Kuaishou upgraded its AI model, launching the Keling 2.5 Turbo model, which significantly enhances video generation capabilities [1] - The new model improves text understanding and response, allowing for a deeper comprehension of complex instructions with causal relationships, enabling refined control over creative expressions [1] - The dynamic effects have been enhanced, allowing for greater motion and camera movement, resulting in smoother and more stable video presentations [1] - The Keling 2.5 Turbo model generates a 5-second video in high-quality mode (1080p) for only 25 inspiration points, which is nearly 30% cheaper than the previous 2.1 model at the same level [1] Group 2: Market Outlook - Goldman Sachs forecasts that Kuaishou's new AI model will continue to drive user engagement, maintaining a prediction of AI-related revenue reaching USD 154 million for the fiscal year 2025, compared to the company's guidance of USD 125 million [1] - Goldman Sachs reiterated a "Buy" rating, with a target price of HKD 83, based on an estimated earnings multiple of approximately 14 times for the end of 2026 [1] - The firm anticipates that the new model will provide top-tier performance in AI video generation at a more competitive cost, highlighting that the AI video generation industry is still in its early stages, with rapidly growing market potential that could disrupt traditional advertising and short film production models [1]

智通财经APP获悉，快手-W(01024)涨近4%，截至发稿，涨3.78%，报76.85港元，成交额12.26亿港元。 消息面上，9月23日，可灵AI基座模型再升级，视频生成推出可灵2.5 Turbo 模型，效果进一步提升。文 本理解与响应显著提升，能更深入理解具有因果关系的复杂指令，实现对创意表达的精细化控制。动态 效果进一步增强，能生成动态幅度更大的动作和运镜，在提升运动呈现幅度和强度的同时，更流畅，更 稳定。同时，可灵2.5 Turbo 模型高品质模式(1080p)下生成5秒长度的视频仅需25灵感值，相比2.1模型 同档位便宜近30%。 高盛发研报称，预期快手可灵AI的新模型将推动快手用户参与度持续提升，现时维持快手2025财年AI 相关收入达到1.54亿美元的预测(相对公司指引为1.25亿美元)，并重申"买入"评级，基于2026年底每股 盈利预测的约14倍，给予目标价83港元。高盛预期，可灵新模型能以更具竞争力的成本提供顶级性能的 AI影片生成方案，认为目前AI影片生成行业仍处于早期阶段，市场规模正快速增长，并可能颠覆传统 广告及短剧制作模式，而关键正在于持续升级的模型能力与用户渗透率。 ...

高盛预期，可灵新模型能以更具竞争力的成本提供顶级性能的AI影片生成方案，认为目前AI影片生成 行业仍处于早期阶段，市场规模正快速增长，并可能颠覆传统广告及短剧制作模式，而关键正在于持续 升级的模型能力与用户渗透率。 快手昨日(23日)发布可灵AI2.5Turbo新模型，该模型在文本回应、动画效果、风格一致性及美学质量等 多个维度的能力都显著提升，并进一步增强影片生成的可控性、稳定性与一致性，在高品质模式下生成 5秒影片的定价仅需25点，较同级别的2.1模型降低近30%。 高盛发布研报称，预期快手-W(01024)可灵AI的新模型将推动快手用户参与度持续提升，现时维持快手 2025财年AI相关收入达到1.54亿美元的预测(相对公司指引为1.25亿美元)，并重申"买入"评级，基于 2026年底每股盈利预测的约14倍，给予目标价83港元。 ...

高盛预期，可灵新模型能以更具竞争力的成本提供顶级性能的AI影片生成方案，认为目前AI影片生成 行业仍处于早期阶段，市场规模正快速增长，并可能颠覆传统广告及短剧制作模式，而关键正在于持续 升级的模型能力与用户渗透率。 智通财经APP获悉，高盛发布研报称，预期快手-W(01024)可灵AI的新模型将推动快手用户参与度持续 提升，现时维持快手2025财年AI相关收入达到1.54亿美元的预测(相对公司指引为1.25亿美元)，并重 申"买入"评级，基于2026年底每股盈利预测的约14倍，给予目标价83港元。 快手昨日(23日)发布可灵AI 2.5 Turbo新模型，该模型在文本回应、动画效果、风格一致性及美学质量等 多个维度的能力都显著提升，并进一步增强影片生成的可控性、稳定性与一致性，在高品质模式下生成 5秒影片的定价仅需25点，较同级别的2.1模型降低近30%。 ...

相关事件 快手-W(01024.HK)9月22日耗资7380万港元回购100万股 快手-W(01024.HK)：利润率新高 可灵持续超预 期 特别派息20亿港币 高盛发表研究报告指，快手昨日发布可灵AI 2.5 Turbo新模型，该模型在文本回应、动画效果、风格一 致性及美学质量等多个维度的能力都显著提升，并进一步增强影片生成的可控性、稳定性与一致性。高 盛预期，可灵新模型能以更具竞争力的成本提供顶级性能的AI影片生成方案，认为目前AI影片生成行 业仍处于早期阶段，市场规模正快速增长，并可能颠覆传统广告及短剧制作模式，而关键正正在于持续 升级的模型能力与用户渗透率。 高盛预期可灵AI的新模型将推动快手用户参与度持续提升，现时维持快手2025财年AI相关收入达到1.54 亿美元的预测(相对公司指引为1.25亿美元)，并重申"买入"评级，基于2026年底每股盈利预测的约14 倍，给予目标价83港元。 ...

Core Insights - Kuaishou has introduced OneSearch, an end-to-end generative framework for e-commerce search, addressing challenges in traditional search architectures [1][2][4] Group 1: Innovation and Technology - OneSearch integrates three innovations: Keyword Enhanced Hierarchical Quantization Encoding (KHQE), Multi-Perspective User Behavior Sequence Injection Strategy, and Preference-Aware Reward System (PARS) [2] - KHQE uses RQ-OPQ encoding to model product features, creating a "smart identity" for each product, enhancing retrieval accuracy [2] - The Multi-Perspective User Behavior Sequence Injection Strategy captures both recent preferences and long-term interests, improving user intent understanding [2] - PARS combines multi-stage supervised fine-tuning with adaptive reinforcement learning to capture fine-grained user preference signals, enhancing ranking performance while ensuring diversity [2] Group 2: Performance Metrics - OneSearch has shown significant improvements in various metrics compared to traditional systems, with a 3.22% increase in order volume and a 2.4% growth in the number of buyers [4] - In offline experiments, OneSearch outperformed existing systems in terms of Click-Through Rate (CTR) and Conversion Rate (CVR), with notable improvements in user satisfaction and item quality [5][6] - The system achieved an 8-fold increase in machine computation efficiency and a 75.40% reduction in online inference costs, optimizing resource utilization [5] Group 3: Market Impact and Future Directions - OneSearch's deployment marks a significant breakthrough in replacing traditional search links with generative models in large-scale industrial scenarios [4][6] - The system excels in cold start scenarios, effectively addressing challenges related to long-tail users and newly listed products [6] - Kuaishou plans to explore online real-time encoding solutions and enhance reinforcement learning mechanisms to better match user preferences, aiming for a more intelligent and precise e-commerce search experience [6]

Group 1 - The National Internet Information Office has conducted interviews with platforms including Toutiao and UC Browser due to the presence of inappropriate content in their trending search lists [1] - UC Browser has expressed its commitment to addressing the issues raised by the regulatory body and has established a special task force to enhance the management of its trending search list [1] - UC Browser aims to improve the compliance and responsibility of its trending search list management, ensuring a clearer online environment [1] Group 2 - Previous platforms such as Xiaohongshu, Weibo, and Kuaishou have also responded to regulatory notifications, indicating their commitment to rectify the identified issues [2] - Weibo plans to continuously optimize its trending search algorithm to reinforce the public nature and correct value orientation of its trending list [2] - Kuaishou and Xiaohongshu have stated their intentions to further improve their trending search list management mechanisms and strengthen their content management responsibilities [2]

Core Insights - The article discusses the introduction of OneSearch by Kuaishou, an innovative end-to-end generative framework for e-commerce search, aimed at addressing the limitations of traditional cascading search architectures [1][2]. Group 1: Challenges in Traditional E-commerce Search - Traditional e-commerce platforms utilize a "recall, rough ranking, fine ranking" cascading search structure, which, while stable, faces issues such as chaotic product descriptions, relevance problems, bottlenecks in the cascading structure, and cold start challenges [1]. Group 2: OneSearch Framework Innovations - OneSearch integrates three major innovations: - Keyword Hierarchical Quantization Encoding (KHQE) module, which models product features in both vertical and horizontal dimensions, creating a rich semantic "smart ID" for each product, enhancing retrieval accuracy [2]. - Multi-perspective User Behavior Sequence Injection Strategy, which captures both recent preferences and long-term interests, allowing for a deeper understanding of user intent and improving personalized search accuracy [2]. - Preference-Aware Reward System (PARS), which combines multi-stage supervised fine-tuning with adaptive reward reinforcement learning to capture fine-grained user preference signals, enhancing ranking performance while ensuring diversity in generated results [2]. Group 3: Performance Metrics and Results - Offline experiments indicate that OneSearch significantly outperforms existing cascading systems across various metrics. Online deployment results show a 3.22% increase in order volume and a 2.4% growth in the number of buyers, marking a significant achievement in large-scale industrial applications [4]. - In a comparative analysis, OneSearch demonstrated improvements in metrics such as Click-Through Rate (CTR) and Conversion Rate (CVR), with notable enhancements in overall page satisfaction and product quality compared to traditional systems [5]. - OneSearch also excelled in cold start scenarios, effectively addressing the ranking challenges for long-tail users and newly listed products, indicating its robustness in diverse conditions [6]. Group 4: Future Directions - Kuaishou plans to continue exploring online real-time encoding solutions to bridge the gap between predefined encoding and streaming training, while also integrating more powerful reinforcement learning mechanisms to better match user preferences [6]. - The ongoing technological advancements are expected to lead to a more intelligent, precise, and user-centric e-commerce search experience, fulfilling the ideal of "one-step" search for users [6].

9月19日晚，一则颇具冲击力的消息传来，据市场监管总局官网发布公告，根据前期核查，依法对成都快购科技有限公司涉嫌违反《中华人民共和国电子 商务法》等法律法规的行为立案调查。 天眼查信息显示，成都快购科技有限公司是快手旗下的全资子公司，更是快手电商业务的核心载体，快手电商交易平台"快手小店"便由其运营，负责供货 商入驻、服务商入驻、直播基地入驻等业务。 其法定代表人、总经理杨远熙，监事银鑫，两人均为快手集团的核心高管。 在官方通报中，监管部门对当前直播电商领域的问题提出了严厉批评，点出行业内虚假营销、假冒伪劣等违法违规行为，句句直指快手平台运营的核心问 题。 一直以来，这样的行业顽疾，在快手身上似乎体现得更为突出。作为以下沉市场为核心用户基本盘的平台，快手的用户群体对商品信息的辨别能力相对有 限，信息差的存在，在某种程度上为虚假营销提供了滋生的土壤；而头部主播的"退网"风波以及此前多次违规带货被处罚，也让平台失去了关键的流量锚 点。 据《奇点湃》观察，另一个不容忽视的现象是，在快手的销售排行榜上，缺乏品牌背书、品控标准模糊的白牌商品长期占据C位，成为消费者投诉的重灾 区，让快手电商业务的口碑持续承压。 再向前回 ...