刚刚！DeepSeek，突传重磅！

Core Viewpoint - The anticipation for DeepSeek's R2 model continues, but the company has instead released a minor upgrade for the R1 model, which can now be tested on various platforms. The specifics of the upgrade have not been disclosed, and the R2 model's rumored specifications remain unconfirmed [1][2][3]. Summary by Sections DeepSeek R1 Upgrade - DeepSeek has completed a "small-scale trial upgrade" for the R1 model, inviting users to test it on their official platforms. The API interface and usage methods remain unchanged [2]. - The R1 model, launched in January, has outperformed Western competitors on several standardized metrics while reportedly costing only a few million dollars, leading to a significant drop in global tech stocks as investors reassess the need for large investments in AI services [2]. Anticipation for R2 Model - There are ongoing rumors regarding the R2 model, suggesting it may have a staggering parameter scale of 1.2 trillion, nearly double that of R1's 671 billion parameters. This would place it on par with leading models like GPT-4Turbo and Google's Gemini 2.0 Pro [3]. - The dynamic activation parameters for R2 are reported to be 78 billion, with actual computational consumption being only 6.5% of the total parameters, which could significantly reduce operational costs while maintaining high performance [3]. Industry Expansion and Applications - Despite a decline in DeepSeek's popularity, its technology continues to expand across various industries. For instance, Tibet University has launched the "Tibet Smart Language" platform using DeepSeek, and the local government has deployed the model for smart governance applications [4]. - Companies like Luckin Coffee and Zhongke Jiangnan are integrating DeepSeek's technology into their operations, enhancing customer interaction and financial services through AI [4]. Broader Implications and Market Trends - A recent article highlights DeepSeek's milestone in achieving efficient computation and low-cost training, which could reshape global AI competition by promoting open-source strategies and breaking the Western monopoly [5]. - The article emphasizes the need for China to overcome geopolitical and technological challenges to leverage opportunities in the AI sector, suggesting that the future of AI in China depends on technological resilience and strategic upgrades [5]. Investment Recommendations - In light of global uncertainties, sectors such as commercial aerospace, AI, new energy, and quantum technology are recommended for investment due to their innovation capabilities and supply chain security. The "DeepSeek moment" could catalyze growth in these areas [6].