21世纪经济报道记者 董静怡 近日，有报道称AI大模型公司阶跃星辰考虑在港交所IPO，计划筹集约5亿美元。21世纪经济报道记者 就此消息向阶跃星辰核实，截至发稿暂未收到回复。 此时距离这家公司完成超50亿元B+轮融资刷新行业融资纪录，仅仅过去一个月。 也是一个月前，旷视科技联合创始人、千里科技董事长印奇正式出任阶跃星辰董事长，与CEO姜大昕、 首席科学家张祥雨、CTO朱亦博组成全新核心管理团队。印奇给自己的定位是：负责战略方向与技术方 向制定，抓组织变革与技术攻坚，以及他更擅长的那部分，终端商业化。 从融资、印奇挂帅到IPO传闻传出，都在这一个月内。这家被视为大模型"六小虎"中行事相对低调的公 司，突然按下了加速键。 阶跃星辰的技术路径，一直带有鲜明的创始人烙印。 在2025年，公司的业务方向更加聚焦，将落地重心聚焦于为智能终端设备打造AI智能体（Agent），重 点布局汽车、手机、物联网设备等关键应用场景。数据显示，截至2025年年底，阶跃星辰的终端智能体 的API调用量连续三个季度增长近170%。 CEO姜大昕出身微软，是典型的技术派，信奉"多模态是通往AGI的必经之路"。公司成立仅两年，便构 建起覆盖语 ...

智能体方面，2026年初，名为OpenClaw的开源智能体工具在开发者社区和各大技术论坛引起较大轰 动，它不再是过去那个只会陪聊的聊天机器人，而是进化成了能够接管电脑、协助人类完成具体办公任 务的"办公搭子"，初步展示了生产力大变革的形象。多模态方面，字节发布的SeeDance2.0让大模型具 有了影视行业的生产力，从此前"生成一段画面"走向"完成一个作品"，生成15秒视频的可用率从此前 20%提升至90%，提升效率降低成本，有望推动漫剧等行业进入规模化发展阶段。 中银国际认为，随着大模型能力提升，这两条演进路径逐步清晰：Agent通过理解用户意图、拆解复杂 任务，搭载MCP和Skills等工具，快速覆盖商业办公、法律和金融领域应用；多模态有望通过生成仿真 数据，进一步用于加速具身智能等场景。 算力涨价成为新迹象，凸显算力供应瓶颈，算力产业链有望持续受益 中银国际发布研报称，2026年春节前后，国内外主要AI模型纷纷完成重大升级。随着大模型能力提 升，智能体与多模态两条演进路径逐步清晰。与此同时，智谱(02513)节前发布GLM Coding Plan价格调 整函，这一涨价预示着优秀的大模型企业不缺需求，有望 ...

Group 1: Core Insights - The large model industry is evolving towards reasoning and multi-modality, emphasizing the feasibility of commercial application and improving ROI efficiency [1] - The new Qwen3.5-Plus model from Alibaba has significantly improved reasoning efficiency, with a total parameter count of 397 billion and a 60% reduction in deployment memory usage [2] - During the Spring Festival, the DAU of Qianwen reached 73.52 million, indicating a competitive edge in capturing C-end traffic [3] Group 2: Industry Trends - Since 2025, the development of large models has shown trends such as evolution from "generation" to "reasoning," emphasis on native multi-modality, and a focus on commercial scene implementation and ROI returns [1] - The AI application landscape is shifting from "buying traffic" to "buying mindset" and "buying pathways" to achieve customer retention [3] Group 3: Recommended Stocks - Companies to focus on include Zhizhu (02513), MiniMax-WP (00100), and Alibaba-W (09988) [2] - Additional recommendations include Cao Cao Travel (02643), SF Express City (09699), Damai Entertainment (01060), Focus Media (002027), and Bilibili-W (09626) [2]

Summary of Conference Call Notes Industry Overview - The conference call discusses the AI large model industry, highlighting significant developments in February with the release of 17 major models by tech giants in China and the US, including OpenAI's GPT-5.3-Codex, Google's Gemini 3.1 Pro, and others from domestic companies like Zhizhu and MiniMax [1][2]. Key Points and Arguments 1. **Model Releases and Market Dynamics** - February saw the launch of 17 significant large models, indicating a rapid iteration in AI technology by major players [1]. - Zhizhu's GLM-5 Coding Plan sold out immediately, with a price increase of 30% within 20 days, showcasing strong market demand [1]. - MiniMax's M2.5 achieved over 10,000 expert builds globally within a day of its open-source release [1]. 2. **Technological Advancements** - Zhizhu has initiated a computing power partner recruitment plan, achieving deep adaptation with domestic chips like Huawei Ascend and others [2]. - The focus on agent capabilities is becoming a core direction for model iteration, with Claude Opus 4.6 introducing multi-agent collaboration and OpenAI Codex scoring 57% on SWE-Bench Pro [2]. - MiniMax M2.5 surpassed Claude Opus in multi-language complex environments, indicating a competitive edge [2]. 3. **Market Concerns and Responses** - There are concerns about accelerated model iterations leading to computing power bottlenecks and commercialization challenges; however, the adaptation of domestic chips and open-source ecosystems are seen as solutions [3]. - Despite perceptions of domestic models lagging behind international leaders, Zhizhu and MiniMax are reportedly on par or even surpassing in programming and complex task execution [3]. - The commercial viability of video generation tools is questioned, yet Kuaishou and ByteDance have validated consumer willingness to pay, indicating strong demand in B2B sectors like film and advertising [3]. 4. **Investment Opportunities** - The outlook for investment in models and applications is positive for the year [4]. - Recommended companies include Alibaba, Tencent, Kuaishou, and Kunlun Wanwei, with a focus on Zhizhu and MiniMax [5]. - There is a strong interest in agent and programming assistant opportunities, as well as multi-modal and AIGC sectors, with recommendations for Tencent, Bilibili, and others [5]. Additional Important Content - The transition from professional software to natural language interaction in content production tools is noted as a significant paradigm shift [3]. - The price increase of Zhizhu's GLM-5 coding plan reflects the stronger capabilities of large models, leading to greater pricing power and demand shifts [3].

Investment Rating - The industry investment rating is "Positive" (maintained) [4] Core Insights - The development of large models emphasizes input efficiency and commercialization, with intense competition for consumer-facing entry points [1] - The new model "Qwen3.5-Plus" from Qianwen significantly enhances inference efficiency and integrates with Alibaba's ecosystem, potentially increasing traffic and order volume for downstream service providers [2] - During the Spring Festival, AI applications saw a surge in daily active users (DAU), with Qianwen reaching 73.52 million DAU, indicating a competitive landscape for consumer traffic [2] Summary by Sections - **AI Application C-end Positioning**: Qianwen's DAU increased significantly during the Spring Festival, leveraging red envelope activities to capture consumer traffic. Qianwen leads with 3 billion yuan in revenue, while competitors like Doubao and Tencent's Yuanbao also show strong user engagement [2] - **Industry Trends**: Since 2025, the large model industry has shifted towards emphasizing inference and multimodal capabilities, focusing on commercial viability and ROI. Recent developments include significant price adjustments and performance improvements from companies like Zhipu and MiniMax [2] - **Recommended Stocks**: The report suggests investing in Alibaba, with additional attention to companies like Cao Cao Travel, SF Express, Damai Entertainment, Focus Media, and Bilibili-W [2]

Core Insights - The domestic large model industry is experiencing a peak iteration ahead of the Spring Festival, with ByteDance launching the Doubao Model 2.0 series, which aims to enhance execution capabilities for complex real-world tasks [1] - The Doubao 2.0 model is a significant upgrade since its initial release in May 2024, showcasing ByteDance's aggressive AI strategy [1] - Various AI products are rapidly being deployed in Spring Festival scenarios, with ByteDance planning to distribute over 100,000 tech gifts and cash red envelopes during the 2026 CCTV Spring Festival [1] Group 1: Product Launch and Features - Doubao 2.0 includes multiple models: Pro, Lite, and Mini, with the Pro version designed for complex reasoning and agent tasks [2] - In competitive evaluations, Doubao 2.0 Pro achieved gold medal results in various programming and mathematics competitions, outperforming Gemini 3 Pro [2][3] - The model has enhanced instruction-following capabilities, maintaining consistency and control in multi-step tasks [3] Group 2: Market Position and Competition - The launch of Doubao 2.0 coincides with a period of intense competition in the AI model market, with several other companies releasing new models [6] - ByteDance's strategy includes a comprehensive AI layout, with the introduction of Seedance 2.0 and Seedream 5.0 Lite models, which support real-time retrieval capabilities [6][8] - The company aims to leverage the high traffic of the Spring Festival to enhance user engagement and showcase its AI capabilities [8] Group 3: User Engagement and Application - The Doubao 2.0 Pro model is now available on various platforms, including the Doubao App and web versions, with API services also launched [5] - ByteDance's participation in the Spring Festival Gala as the exclusive AI cloud partner aims to integrate traditional culture with AI technology [8] - The "Doubao New Year" campaign encourages user interaction through AI-generated content, enhancing user experience during the festive season [8]

Core Insights - The current AI competition in China is evolving from a focus on chatbot capabilities to a more diversified narrative, with companies aiming for foundational infrastructure in the AI era [2][3] - Major Chinese tech firms are adopting a "Google narrative," emphasizing a full-stack approach that integrates products, models, cloud, and chips, similar to Google's strategy over the past two decades [3][19] - Startups are shifting their focus from chatbots to more defined areas like coding and agent scenarios, aligning with the strategies of companies like Anthropic [20][22] Group 1: Major Tech Firms' Strategies - Chinese tech giants are increasingly aiming to emulate Google's model, with leaders like Baidu and Alibaba emphasizing AI-first strategies and integrated solutions [3][5] - The unique selling point of Google's Gemini lies in its multimodal capabilities, which differentiate it from competitors like ChatGPT and Claude [3][4] - The development of video generation models, such as ByteDance's Seedance 2.0, indicates that Chinese firms are beginning to lead globally in certain AI capabilities [4][5] Group 2: Business Models and Market Dynamics - The AI marketing market is projected to grow from 20.9 billion yuan in 2020 to 53 billion yuan by 2024, with a compound annual growth rate of 26.2% [11] - Different business models are emerging, with some companies focusing on scalable throughput while others target vertical industries for immediate production [9][10] - The integration of multimodal tools is expected to enhance advertising efficiency, as visual content can better support the advertising ecosystem of major tech firms [12][8] Group 3: Startups' Shift in Focus - Startups are moving away from the chatbot model, which has high costs and low retention, towards coding and agent scenarios that offer clearer commercial logic [21][22] - Companies like Anthropic are seen as successful examples of balancing high-intensity R&D with sustainable commercialization, influencing Chinese startups to adopt similar paths [26][27] - The recent performance of companies like Zhizhu and MiniMax, which have seen significant stock price increases after announcing new programming models, reflects the positive market response to this strategic shift [31]

Core Viewpoint - The recent release of Seedance 2.0 by ByteDance has significantly narrowed the gap between its AI models and those of leading companies like Google, with the difference now estimated to be as little as one to two months [62][60]. Group 1: Seedance 2.0 - Seedance 2.0 has generated excitement in the AI community, with many users expressing shock and admiration for its capabilities [64][66]. - The model demonstrates strong instruction-following abilities, effectively understanding complex prompts and generating high-quality video content [71][72]. - Users report that Seedance 2.0 has surpassed previous models in terms of performance, making it suitable for creating professional-quality animations and videos [73][74]. Group 2: Seedream 5.0 Lite - Seedream 5.0 Lite, the latest image model from ByteDance, has improved in two key areas: subject consistency and instruction-following ability [20][78]. - Users have noted that the model generates images with better consistency, reducing the "out-of-place" feeling previously experienced with earlier versions [21][78]. - The model's ability to follow complex instructions has been highlighted as a significant advancement, making it easier for users to edit images effectively [82]. Group 3: Doubao Model 2.0 - Doubao Model 2.0 has shown significant improvements in complex reasoning and agent tasks, outperforming its predecessor by a considerable margin [26][83]. - The model is designed to handle multi-modal tasks natively, integrating text, images, and video without the need for separate plugins, which enhances its efficiency [31][87]. - Doubao 2.0 has also reduced inference costs significantly, making it more accessible for commercial applications, particularly in agent scenarios where token consumption is high [45][99]. Group 4: Strategic Positioning - ByteDance's approach to AI development closely resembles that of Google, focusing on integrating models with applications to create a feedback loop that informs future model improvements [100][104]. - The company leverages its large user base and content creators to identify gaps in capabilities, allowing for targeted enhancements in its AI models [102][103]. - The synergy between model development and cloud services, particularly through Volcano Engine, positions ByteDance favorably in the competitive landscape [108][109].

Core Insights - 2026 is anticipated to be a pivotal year for AI, with significant advancements and competition among major players like ByteDance, OpenAI, and Anthropic [1][2] - The launch of new AI models, including ByteDance's Doubao 2.0 and Seedance 2.0, marks a substantial leap in capabilities, particularly in multi-modal understanding and video generation [3][4] Group 1: AI Model Developments - Anthropic and OpenAI have released new foundational models, leading to significant market reactions and a loss of nearly a trillion dollars in market value for major companies [2] - ByteDance's Doubao 2.0 is a multi-modal agent model that has achieved significant improvements in multi-modal understanding, enterprise-level agent capabilities, and reasoning abilities [5][6][12] - Doubao 2.0 has outperformed competitors in various benchmarks, including math and visual reasoning, achieving top scores in multiple assessments [9][10][14] Group 2: Seedance 2.0 and Video Generation - Seedance 2.0 has gained widespread popularity, showcasing its ability to create high-quality videos from text prompts, with notable examples including the adaptation of a short sci-fi story [44][53] - The model supports mixed-modal inputs, allowing users to combine images, videos, audio, and text for video generation, significantly enhancing creative possibilities [56] - Seedance 2.0's video generation capabilities are considered industry-leading, with improvements in realism, physical accuracy, and narrative control [57][60] Group 3: Competitive Landscape - The AI landscape is becoming increasingly competitive, with ByteDance positioning itself alongside major players like OpenAI and Google, particularly in the fields of image and video generation [61][73] - The advancements in AI technology are transforming the upcoming Spring Festival into a battleground for technological innovation rather than just a peak in user traffic [68][74] - The comprehensive technological advancements across various AI domains, including speech and robotics, provide ByteDance with the confidence to compete on a global scale [70][73]

Core Insights - ByteDance has recently launched its latest video generation model, Seedance 2.0, which represents a significant advancement in AI video technology [1][23]. Group 1: Seedance 2.0 Features and Impact - Seedance 2.0 achieves a qualitative transformation of AI video from a "toy" to a "tool," addressing issues like character consistency and style shifts in long videos, ensuring high narrative coherence [3][24]. - The model supports multi-modal input (images, videos, audio, text) and allows precise control over each material's use, making the creative process more intuitive [3][24]. - Enhanced physical simulation reduces "AI twitching" phenomena, improving the fluidity and realism of generated actions, while achieving millisecond-level synchronization of audio and visuals [3][24]. Group 2: Cost Reduction and Market Opportunities - Seedance 2.0 is expected to significantly lower content production costs, particularly benefiting the short video and micro-drama sectors, with a projected increase in AIGC content market share [4][25]. - The AI-generated content market is anticipated to grow rapidly, with short videos and user-generated content (UGC) driving platform growth [4][25]. - AI short dramas are poised for explosive growth, with AI real-time commentary dramas expected to dominate, benefiting from high narrative efficiency and low production costs [5][26]. Group 3: Market Projections and Financial Implications - The Chinese animated drama market is projected to grow from 168 billion yuan in 2025 to 243.6 billion yuan in 2026, marking a 45% increase, with AI-generated dramas capturing a significant market share [6][27]. - AI drama production costs have been reduced from 1,500-4,000 yuan per minute to approximately 100-300 yuan per minute, with production cycles shortened from 30-45 days to 7-10 days [8][29]. - The profitability of AI dramas is confirmed, with net profits for high-viewership works reaching 200,000-300,000 yuan under paid models [8][29]. Group 4: Industry Dynamics and Competitive Landscape - The AI drama industry is expected to see a shift towards high-quality, serialized content, with Seedance 2.0 acting as a catalyst for this transformation [9][32]. - The industry value chain includes AI technology providers, content generation teams, and distribution platforms, with quality IP being crucial for competitive advantage [9][32]. - The global market for professional creators in video production is estimated at $120 billion, indicating substantial growth potential for companies like ByteDance and Kuaishou [17][43].