月之暗面开源多模态Kimi-2506 金十数据6月23日讯，大模型平台月之暗面（MoonshotAI）对其开源的多模态模型Kimi-VL-A3B- Thinking进行了大升级，发布了2506版本。在性能表现上，Kimi-VL-A3B-Thinking-2506实现了更聪明且 更省token的突破。在多模态推理基准测试中取得了更好的准确性：MathVision上达到56.9（提升 20.1），MathVista上为80.1（提升8.4），MMMU-Pro上是46.3（提升3.2），MMMU上为64.0（提升 2.1），同时平均所需的思考长度减少了20%。 （AIGC开放社区） ...

Core Viewpoint - The article discusses the performance of Xiaomi's MiMo-VL-7B multi-modal model, highlighting its strengths and weaknesses compared to the Qwen2.5-VL model, particularly in various testing scenarios. Group 1 - MiMo-VL-7B model outperforms several multi-modal understanding models, especially Qwen2.5-VL, in various tests [3][5]. - The testing results indicate that the SFT (Supervised Fine-Tuning) and RL (Reinforcement Learning) versions of MiMo-VL-7B show similar performance, while the "think" version significantly outperforms the "no-think" version [5][6]. - MiMo-VL-7B's performance in recognizing handwritten OCR is noted to be poor [5][9]. Group 2 - In table recognition tasks, MiMo-VL-7B's "think" model performs well, while the "no-think" model and Qwen2.5-VL struggle [9][10]. - For medium complexity tables, MiMo-VL-7B-SFT "think" model approaches correctness, while other models fail [18][19]. - The article emphasizes that MiMo-VL-7B-SFT "think" model shows better results in complex table recognition compared to its counterparts [26][27]. Group 3 - The article concludes that Xiaomi's MiMo-VL model is impressive overall, particularly the "think" model, which excels in most capabilities except for handwritten OCR [67][68]. - Despite its strengths, the article suggests that the claims of MiMo-VL-7B significantly outperforming the 72B model may be exaggerated [68].

Investment Rating - The report maintains an "Increase" rating for the media industry, indicating a positive outlook for the sector [6]. Core Insights - The media sector has shown a 1.38% increase during the week of June 9-13, driven by themes such as new consumption [10][18]. - Key areas of growth for 2025 include AI applications, IP monetization, and mergers and acquisitions, with a focus on multi-modal industry directions and companies with IP advantages [1][18]. - The report highlights the upcoming summer film season in 2025, with over 60 films scheduled for release, including a diverse range of genres [2][20]. - ByteDance's release of the Doubao model 1.6, a leading multi-modal model, marks a significant advancement in AI capabilities within the industry [3][20]. Summary by Sections Market Overview - The media sector's performance is buoyed by new consumption trends, with a notable increase in stock prices for companies like Yuanlong Yatu and Chuanwang Media [10][13]. - The report identifies the top-performing stocks in the media sector, with Yuanlong Yatu leading at a 42.9% increase [13][16]. Sub-sector Insights - **Resource Integration**: Companies such as China Vision Media and Guangxi Broadcasting are highlighted for their potential in resource consolidation [18]. - **AI Focus**: Companies like Rongxin Culture and Aofei Entertainment are noted for their advancements in AI applications [18]. - **Gaming Sector**: Strong recommendations are made for companies with solid performance, including Shenzhou Taiyue and Giant Network [18]. - **State-owned Enterprises**: Companies like Ciweng Media and Anhui New Media are emphasized for their growth potential [18]. - **Education Sector**: Xueda Education is mentioned as a key player in the education sub-sector [18]. Key Events Recap - The report discusses the launch of the "China Film Consumption Year" initiative, aimed at boosting audience engagement during the summer film season [20]. - The performance of the domestic film market is highlighted, with significant box office figures reported for recent releases [22][24]. Data Tracking - The report provides insights into the gaming sector, noting popular upcoming games and their expected impact on the market [21]. - It also tracks viewership data for television series and variety shows, indicating audience preferences and trends [25][26].

Market Performance - On June 9, the Hong Kong stock market opened higher, with the Hang Seng Index rising over 1% and the Hang Seng Tech Index increasing by 2.33% [1] - The Hang Seng Tech ETF (513580) saw an intraday increase of 2.82%, with notable gains in stocks such as Kingdee International (up over 6%), Tencent Music, Meituan, and JD.com [1] - The Hang Seng Internet ETF (159688) surged by 3.77% [1] Upcoming Events - Citic Securities highlighted that ByteDance will hold the Volcano Engine Force Conference on June 11 in Beijing, featuring comprehensive upgrades to the Doubao model family and multiple sub-forums covering AI technology innovations and industry applications [1] - The main forum on June 11 will include product launches and discussions on AI Coding and AI Agent, with industry-specific forums focusing on AI applications in finance, automotive, ecology, and healthcare [1] - June 12 will be dedicated to developer exchange, with participation from various corporate partners across sectors like chips, automotive, smart terminals, and software applications [1] Industry Trends - Citic Securities noted a recent surge in multimodal dynamic updates, including Google's launch of the Veo 3 video generation model and Doubao's introduction of video call functionality [2] - Kuaishou announced that its AI ARR is expected to exceed $100 million by March 2025, with monthly payments surpassing 100 million RMB in April and May [2] - Tianfeng Securities suggested that investment strategies should focus on three areas: breakthroughs in Deepseek and open-source AI, valuation recovery in consumer stocks, and the continued rise of undervalued dividends [2]

Core Insights - Citic Construction Investment (中信建投) highlights the recent surge in multimodal dynamic updates in the AI sector, indicating a significant trend towards enhanced video generation and communication technologies [1] Group 1: Company Developments - On May 21, Google officially launched the Veo3 video generation model at the 2025 I/O conference, achieving AI video and audio synchronization [1] - On May 23, Doubao introduced a video call feature that supports real-time video communication and screen sharing [1] - Kuaishou announced that its AI ARR is expected to exceed 100 million USD by March 2025, with monthly payment amounts surpassing 100 million RMB in April and May [1] Group 2: Industry Trends - The upcoming Apple WWDC 2025 on June 10 and ByteDance's Force 2025 conference on June 11 are anticipated to accelerate the deployment of multimodal models and edge AI products [1]

Summary of Conference Call Notes Industry Overview - The conference call primarily discusses the **cloud services industry** in North America, focusing on the performance of major players such as **Microsoft, Meta, Google, and Amazon** [1][2][3]. - The **optical module sector** is highlighted as experiencing strong and sustained demand, indicating a long-term growth trend rather than a short-term rebound [1][7]. Key Points and Arguments - **Capital Expenditure (CAPEX) Adjustments**: - Initial CAPEX forecasts for 2025 were downgraded due to concerns over computational power investments and tariffs, but were later revised upwards to **$320 billion**, reflecting increased market confidence [1][4][3]. - Microsoft and Meta reported revenues and guidance that exceeded expectations, with AI significantly contributing to their performance [1][5]. - **Demand Forecasts**: - Demand for optical modules is expected to grow, with indications that 2026 demand may exceed previous expectations [1][8]. - The market is currently pessimistic about growth in 2026, with predictions of a significant decline in growth rates compared to previous years [9][13]. - **Investment Recommendations**: - Investment strategies should prioritize leading companies such as **宏盛, 旭创, 天孚通信**, and consider **新易盛** due to lower valuations [21]. - Companies like **世嘉光子, 博创科技**, and **新易盛** have reported better-than-expected quarterly results, indicating strong performance in the supply chain [22]. Additional Important Insights - **Technological Innovations**: - The optical communication industry is influenced by emerging technologies such as AI and the metaverse, which are expected to drive development and investment [1][12]. - Breakthroughs in AI model training, particularly in multimodal models, are anticipated to have significant implications for the industry [14]. - **Market Dynamics**: - The cyclical nature of cloud service providers' CAPEX, characterized by three years of double-digit growth followed by a year of low or negative growth, directly impacts the optical communication sector [10][11]. - The entry of new players like **Apple** into the AI space is expected to enhance market demand and optimism [20]. - **Company-Specific Insights**: - **Oracle** and **AIT** are noted for their rapid growth, with expectations of significant market share increases in the coming years [18][19]. - Companies like **德科立** and **源杰科技** are also highlighted for their strong order books, suggesting potential for future performance [23]. Conclusion - The overall sentiment regarding the optical module sector and cloud services industry is cautiously optimistic, with strong demand signals and potential for growth despite some market pessimism regarding 2026 forecasts. Investment strategies should focus on leading firms and monitor emerging technologies that could influence market dynamics.

Summary of Key Points from the Conference Call Company Overview - The conference call focuses on **Kying Network**, a company operating in the gaming industry, particularly in the **legendary game market**. Core Insights and Arguments - The overall valuation of the gaming sector remains between **15-18 times**, with expectations for strong performance in the summer gaming season and AI applications, suggesting investors should overweight the gaming sector [2][4] - Kying Network holds over **50% market share** in the legendary game market, utilizing user platform development and ecosystem creation to extend player lifecycles and reduce marketing costs. The revenue from the "Legend Box" product has significantly increased, with daily active users steadily rising [2][5] - Since Q4 2024, Kying Network has accelerated the launch of new products, including the SLG product "Three Kingdoms: The Return of Hearts," and major IP products like "Monopoly" and "King of Fighters," with multiple launches expected in August and September [2][6] - The company is actively expanding its overseas business, having established offices in Hong Kong and South Korea, and acquired retro IPs. The overseas business is expected to continue its high growth trajectory of **220%** from 2024, with a focus on Southeast Asian markets [2][7] - In the AI sector, Kying Network is developing AI companionship and social applications, with plans to release AI game engine version 2.0 in the summer, and is exploring AI-assisted content creation [2][8] Additional Important Content - The revenue from the "Legend Box" product grew from **200 million** in 2022 to **600 million** in 2023, and is projected to reach **900 million** in 2024, indicating a high gross margin due to its advertising revenue model [5] - Daily active users increased from **400,000** at the beginning of 2024 to **450,000** by the end of the year, with a target of reaching **500,000** [5] - Kying Network's current valuation is **17 times**, with potential advantages through its IP platform and AI initiatives, allowing it to break free from traditional gaming product cycles [3][10] - The company is also exploring AI toys and collaborating with Dapeng Glasses to develop an AI glasses ecosystem [9]

Core Viewpoint - MiniMax is preparing to launch a text reasoning model, codenamed M+, which could significantly impact the company's future and its position in the competitive AI landscape [2][4][25]. Group 1: Upcoming Product Launch - MiniMax has been developing the M+ text reasoning model for over six months and will release it alongside a technical report [2]. - The launch of the M+ model is crucial as it will serve as a benchmark for MiniMax's competitiveness in the AI market, especially after the release of DeepSeek R1 [5][25]. - The company has chosen a hybrid approach by not integrating DeepSeek in its domestic applications while opting for integration in overseas AI applications [3][5]. Group 2: Competitive Landscape - The AI industry is shifting from the "AI Six Little Tigers" narrative to a focus on the "Five Giants" in Silicon Valley, which does not prominently feature MiniMax [5][18]. - MiniMax's delayed entry into the reasoning model market compared to competitors could affect external confidence in the company [4][5]. Group 3: Strategic Moves - MiniMax has made several strategic moves in 2025, including acquiring an AI video startup and rebranding its AI application from "海螺AI" to "MiniMax" [6][9]. - The company is restructuring its product matrix to clearly differentiate between text and video model capabilities [10][11]. Group 4: Organizational Structure - MiniMax's organizational structure includes four main teams focused on text, video, image, and voice models, but its sales team is notably small, comprising only about 3% of the total workforce [13]. - The company adopts a pure API model for B2B clients, which influences its sales strategy and organizational focus [13][14]. Group 5: Financial Performance and Valuation - MiniMax raised $600 million in a Series A round in March 2024, achieving a post-money valuation of $2.5 billion, with indications that its current valuation has exceeded this figure [16]. - The company has engaged in multiple undisclosed funding rounds since then, indicating strong investor interest [16]. Group 6: Commercialization and Market Position - MiniMax's commercial success is primarily driven by its voice model, while the performance of its video model remains less clear [15][27]. - The company is navigating a competitive landscape where the monetization of multimodal models is becoming increasingly important [26][29].

Core Viewpoint - OpenAI's o3 reasoning model has broken traditional boundaries of text-based thinking by integrating images directly into the reasoning process, achieving a new level of multimodal reasoning capabilities [1][4][29] Group 1: Model Capabilities - The o3 model can analyze images and derive answers by focusing on relevant areas, such as formulas in a physics exam or structural elements in architectural drawings, achieving a 95.7% accuracy on the V* Bench visual reasoning benchmark [1] - DeepEyes, developed by a collaboration between Xiaohongshu and Xi'an Jiaotong University, has demonstrated similar capabilities to o3, allowing for reasoning with images without relying on supervised fine-tuning [1][29] Group 2: Reasoning Process - DeepEyes employs a three-step reasoning process: global visual analysis, intelligent tool invocation, and detail reasoning identification, showcasing its ability to think with images [7][10] - The model's architecture introduces a "self-driven visual focus" mechanism, allowing it to dynamically determine when to utilize image information based on the reasoning context [14] Group 3: Learning Mechanism - DeepEyes utilizes an outcome-based reinforcement learning strategy, inspired by biological evolution, to develop its image reasoning capabilities without the need for supervised fine-tuning [18][19] - The learning process is divided into three stages: a novice phase with low accuracy, an exploration phase with increased tool usage, and a mature phase where the model effectively predicts key areas for analysis [21] Group 4: Performance Metrics - DeepEyes has shown superior performance in various visual reasoning tasks, achieving a 90.1% accuracy on the V* Bench and outperforming existing workflow-based methods [23] - The model also exhibits enhanced mathematical reasoning capabilities, indicating its potential for cross-task performance [24] Group 5: Advantages of DeepEyes - Compared to traditional models, DeepEyes offers a simpler training process, stronger generalization capabilities, end-to-end joint optimization, deeper multimodal integration, and inherent tool invocation abilities [26][28][29]

Market Overview - The A-share market saw most major indices decline today, with the Shanghai Composite Index down 0.22% closing at 3380.19 points, the Shenzhen Component down 0.72% at 10219.62 points, the Sci-Tech 50 down 0.48% at 990.71 points, and the ChiNext Index down 0.96% at 2045.57 points [1] - Overall, there were 4451 stocks that declined, while only 882 stocks rose, and 77 stocks remained flat [1] - The total trading volume in the Shanghai and Shenzhen markets was 11027 billion, a decrease of 707.55 billion compared to the previous trading day [1] Industry Performance - Among the Shenwan first-level industries, all sectors except for banking, media, and household appliances experienced declines today [1] - The sectors that led the decline included beauty care, social services, basic chemicals, environmental protection, real estate, and electric equipment [1] Sector Highlights - The top-performing concept sectors today included selected banking, smart speakers, multimodal models, central enterprise banks, ChatGPT, online gaming, K-12 education, selected air transport, Kimi, selected insurance, IGBT, Chinese corpus, short drama games, internet celebrity economy, and central enterprise automobiles [1]