新型大语言模型(LLM)
Search documents
中国AI奋起直追,DeepSeek冲击会再来?
日经中文网· 2025-08-26 08:00
Core Viewpoint - The Chinese stock market is witnessing strong performance in technology stocks, with increasing confidence that Chinese artificial intelligence (AI) can develop independently of American companies like Nvidia. The recent release of DeepSeek's new large language model (LLM) has catalyzed this sentiment [2][4]. Group 1: DeepSeek's Developments - DeepSeek released its new LLM "V3.1" on August 21, which is seen as a transitional model before the anticipated "R2" due to delays in development related to Huawei's semiconductor capabilities [4]. - The market is paying attention to DeepSeek's announcement that V3.1 utilizes UE8M0 FP8 Scale parameters, designed for upcoming domestic chips, indicating a shift towards integrating software and hardware in Chinese AI [4][6]. Group 2: Market Reactions - The Shanghai Stock Exchange's "STAR 50 Index," composed of 50 stocks from the high-tech sector, rose by 8.6% on August 22, reaching its highest level since February 2022, reflecting growing optimism in the AI sector [4]. - The stock price of Cambricon Technology, referred to as the "Chinese Nvidia," has doubled compared to the end of July, indicating strong market interest and potential in domestic AI chip development [5][6]. Group 3: Implications for AI Development - The instability of H20 supply is believed to have contributed to the delays in the development of DeepSeek's R2 model, which has relied on both Huawei and Nvidia chips for training and inference [7]. - DeepSeek's clear shift towards integrating with domestic semiconductors could significantly alter the landscape of AI development in China, potentially reducing reliance on foreign technology [7].