Group 1 - The Hong Kong stock market experienced a strong start to 2026, with the Hang Seng Index rising by 2.76% to 26,338.47 points, and the Hang Seng Tech Index increasing by 4% to 5,736.44 points. The market turnover reached 140.86 billion HKD, showing an increase from the previous trading day [1] - Baidu Group's AI chip company Kunlun has submitted a listing application to the Hong Kong Stock Exchange, with its valuation rising from approximately 13 billion RMB in 2021 to 21 billion RMB by July 2025 [3] - The Hong Kong IPO fundraising amount for 2025 reached 285.69 billion HKD, a significant increase of 224% year-on-year, with 117 companies listed, marking a growth of 67.14% [2] Group 2 - The Turkish government announced that from January 2, 2026, Chinese passport holders will enjoy visa-free travel and transit, allowing a maximum stay of 90 days within any 180-day period [2] - Multiple public fund institutions have released optimistic investment strategies for 2026, with technology being a key focus area. They expect market dynamics to shift from valuation-driven to a dual drive of "profit and valuation" [2] - The European manufacturing PMI for December 2025 was reported at 48.8, below expectations and previous values of 49.2 [8]

Core Insights - The article highlights the emergence of the IQuest-Coder-V1 model series, which has gained significant attention in the tech community for its performance in code generation and understanding tasks [1][2]. Model Performance - The IQuest-Coder-V1 model, particularly the 40B parameter version, achieved an impressive score of 81.4% on the SWE-Bench Verified leaderboard, surpassing models like Claude Opus-4.5 and GPT-5.2, which are speculated to have parameter scales in the hundreds of billions to trillions [2][50]. - The model series includes versions with 7B, 14B, and 40B parameters, each offering Instruct and Thinking variants tailored for different use cases [14][15]. Technical Specifications - The IQuest-Coder-V1 series emphasizes "engineering-friendly" design and long context usability, supporting a maximum context length of 128K tokens and a vocabulary size of 76,800 tokens [22][25]. - The 40B parameter version features a Loop variant that enhances parameter utilization efficiency, achieving significant reductions in HBM and KV Cache overhead while improving throughput [19][20]. Training Methodology - The training strategy, termed "code-flow multi-stage training," focuses on learning from the evolution of code rather than static code snippets, incorporating a triplet data structure to capture changes over a project's lifecycle [38][43]. - This approach allows the model to understand the dynamic evolution of software logic, capturing differences before and after modifications [46][47]. Deployment and Accessibility - The models are designed for deployment on consumer-grade GPUs, with the Int4 version capable of running on a single H20 inference card [53][54]. - The IQuest-Coder series has been open-sourced on platforms like GitHub, making it accessible for developers and researchers [11]. Company Background - IQuest-Coder is developed by Ubiquant Holding Limited (九坤投资), a prominent quantitative investment firm in China, known for its focus on AI and high-frequency trading [57][64]. - The company has established multiple research labs, including an AI Lab, and has a strong team with a high percentage of members holding advanced degrees from top universities [62][64].

Group 1 - The establishment of the Artificial Intelligence Research Promotion Center by the Chinese Academy of Social Sciences aims to address fundamental questions regarding the direction of the intelligent revolution and to provide a "Chinese solution" for global AI governance reform [1] - The center will focus on the intersection of artificial intelligence and major theoretical and practical issues in philosophy and social sciences, striving to create a leading research platform that reflects national will and mission [1] - The development and governance of artificial intelligence are not only technological issues but also involve economic, social, cultural, legal, ethical, diplomatic, and international political aspects [1] Group 2 - Artificial intelligence has become a core force driving industrial transformation and enhancing national competitiveness, with large models rapidly becoming foundational technologies across various application fields [2] - The introduction of large language models into social science research methods is reshaping traditional research approaches, enabling automation and efficiency in text analysis, data generation, and simulation experiments [2] - The emergence of code large models significantly lowers programming barriers, facilitating more efficient data processing and analysis for researchers, thus promoting innovation and development in social science research [2]

Core Insights - aiXcoder's intelligent software development solution has been recognized as an "Outstanding Case in Software R&D" at the TiD 2025 Quality Competitiveness Conference due to its successful application in a state-owned bank, resulting in a 30% increase in overall development efficiency [1] Group 1: Technology Implementation - The solution includes three key technological implementations: 1. Deployment of a code large model trained specifically for code characteristics, enhancing performance in software development scenarios through context-aware code generation, completion, defect fixing, and unit test generation [1] 2. Personalized training for banking-specific code, utilizing the bank's private code and documentation to create a bespoke code large model that aligns with the bank's business logic and coding style, all while maintaining the core model's performance [1] 3. Private deployment that meets strict security requirements, operating entirely within the internal network to ensure data security, optimizing hardware resource usage, and supporting high concurrency scenarios [2] Group 2: Performance Metrics - The proportion of AI-generated code in development has increased from 10% before training to 35% after implementation, with specific scenarios allowing for up to 60% of coding tasks to be assisted by AI [2]

Core Viewpoint - Nvidia has recently open-sourced its code reasoning model, which includes three parameter versions: 32B, 14B, and 7B [1] Group 1 - The open-source model is based on Alibaba's Tongyi Qianwen Qwen2.5 series, specifically Qwen2.5-32B, Qwen2.5-14B, and Qwen2.5-7B [1]