Financial Performance - Revenue increased from RMB 3,405.8 million in 2023 to RMB 3,772.1 million in 2024, representing an increase of approximately 10.7%[8] - Gross profit rose from RMB 1,500.8 million in 2023 to RMB 1,619.7 million in 2024, marking an increase of about 7.9%[8] - Annual loss decreased from RMB 6,494.7 million in 2023 to RMB 4,306.6 million in 2024, a reduction of approximately 33.8%[8] - Adjusted EBITDA loss improved from RMB 4,369.0 million in 2023 to RMB 3,089.2 million in 2024, reflecting a decrease of about 29.3%[8] - In 2024, the total revenue of SenseTime Group increased by 10.8% year-on-year, reaching RMB 3,772.1 million, with generative AI revenue surpassing RMB 2,404.0 million, a year-on-year growth of 103.1%[13] - Generative AI now accounts for 63.7% of the group's total revenue, up from 34.8% in 2023, marking it as the largest revenue contributor[14] - The group's gross profit for 2024 was RMB 1,619.7 million, with a gross margin of 42.9%, while total management and sales expenses decreased by 9.1% year-on-year[13] Customer Engagement and Product Development - The company achieved a significant increase in customer willingness to pay, with order amounts growing sixfold compared to 2023[12] - Monthly user engagement increased eightfold compared to 2023, driven by enhanced user experience and product offerings[12] - The company launched the first domestic large model surpassing GPT 4-Turbo performance in April 2024, followed by a video streaming interactive model in July 2024[12] - The productivity tools based on the multi-modal model have seen a 6x year-on-year increase in customer payment willingness, represented by order amounts in 2024[24] - The user base for office and code assistant products has surpassed 1.5 million, with daily invocation exceeding 1 million times and processing over 3.5 billion tokens daily[26] Research and Development - Research and development expenses increased to RMB 4,131,884 thousand in 2024 from RMB 3,465,766 thousand in 2023, indicating a focus on innovation[40] - R&D expenses grew by 19.2% from RMB 3,465.8 million in 2023 to RMB 4,131.9 million in 2024, driven by investments in training and fine-tuning base models[45] - The company aims to achieve carbon neutrality by 2030 and has implemented measures that saved 800 MWh of electricity in 2024[34] - The company has been recognized as the first to pass the enhanced level assessment of the National Computing Power Service Capability Maturity Model (CPMM)[35] Financial Position and Cash Flow - Total assets increased from RMB 32,888.0 million in 2023 to RMB 34,599.5 million in 2024, an increase of approximately 5.2%[7] - Total liabilities rose from RMB 9,732.6 million in 2023 to RMB 10,957.8 million in 2024, an increase of about 12.6%[7] - As of December 31, 2024, the total cash reserves of the group amounted to RMB 12,752.2 million, and trade receivables increased by 19.0% year-on-year to RMB 4,623 million[13] - The net cash used in operating activities for the year ended December 31, 2024, was RMB (3,926.7) million, compared to RMB (3,234.3) million in 2023, indicating an increase in cash outflow of 21.5%[64] - The net cash generated from financing activities for the year ended December 31, 2024, was RMB 6,259.9 million, significantly higher than RMB 1,083.6 million in 2023, marking a 478.5% increase[67] Strategic Partnerships and Market Expansion - The company has established partnerships with emerging industry clients in areas such as embodied intelligence and AI for science, expanding its market reach[15] - In the smart automotive sector, the delivery of the "Zhenying" system increased by 29.2% year-on-year, with over 1.67 million new deliveries and a cumulative total exceeding 3.6 million vehicles[29] - The company signed strategic cooperation agreements with Dongfeng and Chery to deliver smart driving solutions based on the domestically produced Horizon J6 platform in Q1 2025[30] Governance and Compliance - The company held a total of 7 board meetings, 5 audit committee meetings, 2 remuneration committee meetings, 1 nomination committee meeting, and 3 corporate governance committee meetings during the reporting period[99] - The audit committee held 5 meetings during the reporting period to discuss audit and financial reporting matters with the auditors[106] - The corporate governance committee conducted 3 meetings to review and monitor the company's governance policies and compliance with legal and regulatory requirements[110] - The company has established a whistleblowing mechanism to encourage internal reporting of suspicious activities, with no significant fraud or misconduct events affecting the financial statements for the year ending December 31, 2024[134] Shareholder Engagement and Dividends - The company encourages shareholders to participate in annual general meetings and provides at least 21 days' notice for such meetings[141] - The company did not recommend the distribution of a final dividend for the year ending December 31, 2024, considering the long-term interests of shareholders[156] - The company maintains a website for regular updates and communication with shareholders, including financial reports and announcements[143] Risk Management - The company has established a comprehensive risk management framework to identify and manage compliance risks, ensuring operations comply with applicable laws and regulations[129] - The company has implemented a series of internal procedures to manage operational risks, aiming to control potential losses through identification, measurement, monitoring, and control of these risks[127] - The audit committee has reviewed the company's risk management and internal control systems, deeming them effective and sufficient[136]

4月22日，中国移动香港有限公司与商汤集团有限公司在香港签署合作备忘录。(中国移动香港有限公司 供图) 根据合作备忘录，中国移动香港与商汤科技将联合推出AI+智能视频分析行业解决方案，开发行业定制 化AI视频分析系统，实现即时行为识别与多场景智能监测。双方也将启动垂直行业大模型训练计划， 联合优化AI算法，整合产业伙伴的领域专业数据，打造符合行业合规要求的预训练模型，并支持私有 化部署的轻量化模型方案。 中新网香港4月22日电 (记者 戴小橦)中国移动香港有限公司(简称"中国移动香港")与商汤集团有限公司 (简称"商汤科技")22日签署合作备忘录。双方表示，将发挥在人工智能(AI)与通信科技领域的核心优 势，助力推动香港发展成为国际创科中心。 商汤科技亚太业务总裁、集团战略规划副总裁史军表示，基于商汤科技领先的视觉AI及大模型能力， 与中国移动香港共同开发创新的应用场景，打造垂域大模型，支持轻量化部署，共同探索人工智能与通 信技术融合的无限可能。未来，商汤科技会继续以"坚持原创"为使命，加强AI技术与各行业的融合，推 动香港创科生态圈发展。(完) 中国移动香港政企客户部总裁张炜表示，此次合作将聚焦AI+智能视 ...

中金公司 AI 产业动态更新：Agent 密集发布、MCP 生态 快速繁荣 2025042120250416 摘要 • OpenAI 发布 O3 和 O4 mini 系列模型，结合图片推理能力，内置联网搜 索、文档解析、图片生成等功能，虽未引起轰动，但展示了其在 AI 技术上 的持续投入。Sora 更新中文生图功能具备良好的指定遵循和风格切换能力。 • 谷歌在 Google Cloud Next 大会上推出 Gemini 2.5 系列推理模型，具 备 Hybrid reasoning 能力，并推出 agent-to-agent 协议以促进协作。 同时，谷歌还更新了视频、语言、音乐生成及图片编辑功能，并与 Google Workspace 深度集成，提升企业级产品能力。 • Meta 发布 LLAMA4，作为全球开源社区广泛使用的基础模型，其革新为 社区带来显著进步。LLAMA4 有三个版本，其中最大的版本仍在训练中。 Maverick 版本表现不错，但存在争议，总体展现出强大的工具调用能力、 高速度及性价比。 • 商汤科技发布 SenseNova V6 系列模型，具有超长思维链，支持图文多 模态推理能力，与阿 ...

- The report introduces the Llama 4 model series, which includes Llama 4 Scout, Llama 4 Maverick, and Llama 4 Behemoth, highlighting their advanced multimodal capabilities and efficiency through the MoE (Mixture of Experts) architecture[10][11][12] - Llama 4 Scout features 16 experts with 17 billion activated parameters, supports a 10M context window, and is optimized for single H100 GPU deployment, achieving state-of-the-art (SOTA) performance in various benchmarks[11][12] - Llama 4 Maverick employs 128 routed experts and a shared expert, activating only a subset of total parameters during inference, which reduces service costs and latency. It also incorporates post-training strategies like lightweight SFT, online RL, and DPO to balance model intelligence and conversational ability[12][14] - The CoDA method is introduced to mitigate hallucination in large language models (LLMs) by identifying overshadowed knowledge through mutual information calculations and suppressing dominant knowledge biases. This method significantly improves factual accuracy across datasets like MemoTrap, NQ-Swap, and Overshadow[23][25][29] - The KG-SFT framework enhances knowledge manipulation in LLMs by integrating external knowledge graphs. It includes components like Extractor (NER and BM25 for entity and triple extraction), Generator (HITS algorithm for generating explanatory text), and Detector (NLI models for detecting knowledge conflicts). KG-SFT demonstrates superior performance, especially in low-data scenarios, with a 14% accuracy improvement in English datasets[45][47][52] - DeepCoder-14B-Preview, an open-source code reasoning model, achieves competitive performance with only 14 billion parameters. It utilizes GRPO+ for stable training, iterative context length extension, and the verl-pipeline for efficient reinforcement learning. The model achieves a Pass@1 accuracy of 60.6% on LiveCodeBench and a Codeforces score of 1936, placing it in the 95.3rd percentile[53][61][64]

一上台，商汤科技董事长兼CEO 徐立就感叹，"如果三个月不更新自己的认知，可能就会被淘汰。" 4月10日，商汤举办2025技术交流日，徐立正式发布全新升级的"日日新SenseNova V6"（以下简称"日日 新V6"）大模型体系。 在徐立看来，多模态模型和通用人工智能的发展，画上约等号，以计算机视觉起家的商汤，从视觉能力 到原生多模态模型的布局，则是自然延伸。 商汤科技联合创始人兼大模型首席科学家林达华向《21CBR》记者表示，公司去年5、6月份就在做多模 态的探索，到了9、10月，技术路线基本跑通。 林达华称，之所以专注多模态推理，而非纯文本赛道的竞争，在于坚信未来的交互，必然是多模态的。 日日新V6，作为拥有超6000亿参数的MoE原生多模态通用大模型，凭借单一模型就可以完成文本、多 模态等各类任务。 其技术能力上的突破，重在四个方面： 长思维链：超过200B高质量多模态长思维链数据，最长64K思维链；数理能力：数据分析能力大幅领先 GPT-4o；推理能力：多模态深度推理国内第一，对标OpenAI o1；全局记忆：率先在国内突破长视频理 解，支持10分钟的视频理解及深度推理。 值得一提的是，长记忆。林达华 ...

Core Insights - The article highlights the launch of SenseNova V6, a new multimodal AI model by SenseTime, which enhances reasoning capabilities and cost efficiency in AI applications [2][4] - The model is designed to support various applications, including humanoid robots, and aims to improve human-robot interaction through advanced perception and reasoning [4][5] - The Chinese multimodal AI market is experiencing rapid growth, with projections indicating a market size of approximately 15 billion RMB in 2024, reflecting a year-on-year increase of about 30% [6] Group 1: Product and Technology - SenseNova V6 features breakthroughs in multimodal long reasoning chains, global memory, and reinforcement learning, significantly outperforming competitors like OpenAI's models [2] - The model supports in-depth analysis of mid-length videos and is positioned as one of the strongest in its category, comparable to Gemini 2.5 Turbo [2] - The technology enables robots to understand gestures, respond to environmental inquiries, and provide a more authentic interaction experience [4] Group 2: Industry Applications - SenseTime's end-to-end solutions for embodied intelligence address high data costs and fragmented toolchains, supporting over 10TB of data aggregation per day [4][5] - The AI supermarket project demonstrates the practical application of group intelligence, showcasing a complete AI development system from model training to inference evaluation [5] - The collaboration between Fourier's humanoid robot GRx and SenseTime's SenseNova V6 Omni enhances the robot's ability to understand complex scenarios through deep integration of various data types [5] Group 3: Market Trends - The demand for embodied intelligence in robotics has increased significantly, driven by technological innovation and the need for advanced data training solutions [6] - The multimodal AI market in China is expected to continue its rapid growth, with forecasts suggesting it will exceed 20 billion RMB by 2025 [6] - Industry competition is intensifying, with a focus on maintaining technological innovation, improving model generalization, and ensuring data security and privacy [7]

Core Viewpoint - The article discusses the evolution of artificial intelligence (AI) from its inception to the current era of large models, highlighting key milestones, technological advancements, and the impact on various industries. Group 1: Birth of Artificial Intelligence (Mid-20th Century) - In 1950, Alan Turing proposed the "Turing Test," defining the philosophical goal of AI [3] - The term "Artificial Intelligence" was first used in 1956 at Dartmouth College, marking the transition from philosophical speculation to applied technology [3] - Early AI systems, like the IBM701, had limited computational power, executing only 16,000 operations per second, which is significantly less than modern devices [3] Group 2: Symbolism and Its Failures (1960-1970) - The 1960s saw the rise of "symbolism," where researchers attempted to simulate human reasoning through rule-based expert systems [4] - The MYCIN system developed in 1976 achieved near-expert accuracy in diagnosing blood infections, demonstrating the commercial value of expert systems [4][5] - The "Fifth Generation Computer Systems" project in Japan, launched in 1982 with an investment of $850 million, aimed to create intelligent computers but ultimately failed due to over-reliance on symbolic methods and hardware limitations [8] Group 3: Rise of Machine Learning (1990s-2000s) - The 1990s marked a shift to machine learning, moving from rule-based systems to data-driven approaches, allowing machines to learn from data rather than relying solely on hard-coded rules [10] - IBM's DeepBlue defeated a chess champion in 1997, showcasing the potential of machine learning in closed tasks [12] - The introduction of Google's PageRank algorithm in 1998 demonstrated the commercial value of data correlation, transforming search engines into profitable ventures [12] Group 4: Deep Learning Revolution (2010s-2020) - The 21st century saw the emergence of deep learning, enabling AI to automatically extract features through multi-layer neural networks [13] - AlphaGo's victory over a world champion in 2016 highlighted the capabilities of deep reinforcement learning [13] - The rapid increase in model parameters from 60,000 in LeNet-5 to 600 million in AlexNet illustrated the exponential growth in AI's capacity to handle complex tasks [14] Group 5: Era of Large Models (2021-Present) - The introduction of large pre-trained models like GPT-3 in 2020 has propelled AI towards general intelligence, showcasing advanced language understanding and generation capabilities [15] - Applications of generative AI have expanded across various fields, including content creation, programming assistance, and image generation, significantly enhancing productivity [16] - The competition between open-source and closed-source models has intensified, with companies like HuggingFace promoting open-source development while others like OpenAI focus on proprietary advancements [17] Group 6: Future Directions and Challenges - The future of AI is expected to focus on specialized models for high-value sectors such as healthcare and finance, emphasizing efficiency and cost-effectiveness [38] - The relationship between AI and human employees is anticipated to evolve into deeper integration, enhancing decision-making and innovation within organizations [38] - Ethical challenges and societal risks associated with AI, such as job displacement and privacy concerns, remain critical issues that need addressing [39]

Core Insights - SenseTime launched the upgraded "Riri Xin V6" large model system in Shanghai on April 10, enhancing multi-modal integration to expand the application scenarios of artificial intelligence [1] - In the era of general artificial intelligence, large models empower humanoid robots with stronger external perception and deep thinking capabilities, while also being widely applied in various everyday life scenarios [1] - The transition from "cloud-based" to "grounded" applications signifies a significant advancement in making AI more accessible and practical for the general public [1]

Summary of the Conference Call on SenseTime Technology Company Overview - **Company**: SenseTime Technology - **Industry**: Artificial Intelligence (AI) Key Points and Arguments Performance and Achievements - SenseTime's "Riri Xin" fusion model ranked first in both SuperCLUE and OpenCompass evaluations, achieving a total score of 18.3, tying with DeepCV3, indicating a significant breakthrough in native fusion modality training [2][4][5] - The company launched the Riri Xin 6.0 version, which constructs over 200 billion high-quality tokens for multi-modal long thinking chain data, achieving a length of 64K, significantly enhancing data analysis capabilities, particularly in vertical industries like finance [2][20] Government Support and Industry Growth - The Shanghai government is heavily supporting the AI industry, with the industry scale expected to exceed 450 billion yuan by the end of 2024, and over 60 generative AI models have been registered with the state [2][7] - SenseTime has developed the SenseCore AI computing platform to provide efficient computing power support for large model research and industrial applications in Shanghai [2][8] Technological Innovations - SenseTime's multi-modal models excel in processing unstructured data, improving efficiency and decision-making in scenarios like financial audits and e-commerce price comparisons [2][24] - The company emphasizes the importance of multi-modal models in achieving general artificial intelligence, as they can enhance learning efficiency and address complex problems [12][67] Future Directions and Applications - SenseTime aims to apply its native modality fusion widely across various scenarios to enhance interaction experiences [6][9] - The company is focused on deepening AI applications in key industries and fostering collaboration with academic institutions to build open platforms [9] Market Position and Competitive Edge - According to a report by Frost & Sullivan, SenseTime ranks first in China's generative AI technology stack market due to its continuous investment in technology innovation and high-performance domestic inference engines [3] Real-World Applications - The multi-modal model has been successfully applied in various fields, including automatic driving and smart healthcare, showcasing its ability to solve complex issues and enhance user experience [2][8][24] - In the e-commerce sector, the model can automatically analyze price information across platforms, providing optimal purchasing suggestions [25][26] Challenges and Opportunities - The rapid growth of multi-modal data presents challenges in data management and processing, necessitating the development of adaptive technologies to optimize performance [19][67] - The company is committed to addressing the challenges of data scarcity in the robotics sector through virtual simulation technologies [68][72] Educational Impact - SenseTime's technology is also being integrated into educational tools, enhancing learning experiences through interactive and immersive methods [50][52] Collaboration and Ecosystem Development - SenseTime collaborates with various partners, including Kirin Software, to develop comprehensive solutions that enhance the domestic AI ecosystem [30][59] Additional Important Content - The company is preparing for the World Artificial Intelligence Conference in 2025, aiming to foster international cooperation and share innovative outcomes [9] - SenseTime's advancements in video editing and AI capabilities are set to revolutionize content creation and enhance user engagement [55][57] This summary encapsulates the key insights from the conference call regarding SenseTime Technology's performance, innovations, market position, and future directions in the AI industry.

Investment Rating - The investment rating for the company is "Outperform the Market" [2] Core Insights - The company has achieved over 100% growth in its generative AI business, which has become its largest revenue source, accounting for 63.7% of total revenue in 2024 [6][8] - The company completed a restructuring of its "1+X" organizational framework, focusing resources on core businesses, particularly generative AI and visual AI [6] - The company is set to release its new model, "Riri Xin 6.0," in Q2 2025, which is expected to significantly enhance multimodal understanding and interaction capabilities [6] - The company maintains a leading position in the visual AI market, with a customer repurchase rate increase of 31 percentage points in 2024 [8] - The company has made significant advancements in its autonomous driving business, with over 1.1 million new designated vehicles added in 2024 [8] Financial Data and Forecast - Revenue is projected to grow from 34.06 billion CNY in 2023 to 73.97 billion CNY by 2027, with a year-on-year growth rate of 27% in 2027 [6][9] - The net profit is expected to improve from a loss of 6.44 billion CNY in 2023 to a profit of 305 million CNY by 2027, reflecting a significant turnaround [6][9] - The gross margin is forecasted to increase from 44.07% in 2023 to 57.82% in 2027 [6][9] - The company’s return on equity (ROE) is expected to turn positive by 2026, reaching 1.33% in 2027 [6][9] Market Performance - The company's stock closed at 1.47 HKD on April 1, 2025, with a market capitalization of 54.401 billion HKD [2] - The stock has experienced a 52-week price range of 0.58 to 2.35 HKD [2] Valuation - The company is assigned a price-to-sales (P/S) ratio of 16-20 times for 2025, indicating a fair value range of 2.15 to 2.69 HKD per share [6][9]