After reviewing the evaluation of MiniMax in the Goldman Sachs report, Goldman Sachs believes that MiniMax has "ARR upside potential and an upward skewed risk reward". The report was adjusted to a buy rating on June 9th, when the stock price was around HKD 500. Goldman Sachs buying logic -Key recommendations for AI model sector in companies with small and medium-sized market capitalization -Asymmetric risk return: Clearly indicate that its ARR upside potential is accompanied by an upwardly skewed risk reward, meaning limited downside and large upside potential -The new model catalysis, Hailuo 3 video model (released in July), larger parameter base models such as M3 (3Q), and full modal/encoding performance will drive ARR acceleration and high growth visibility. -Efficiency and gross profit moat: MiniMax demonstrates the highest cost efficiency and API gross profit margin in each iteration, building a long-term sustainable business model with the potential to achieve greater TAM globally. -Differentiation: Avoiding the pure Chatbot red sea, using to C differentiation (overseas, multimodal)+to B light asset API services to maintain unique competitiveness in the giant battle, in line with Goldman Sachs' optimistic model of income quality price ratio path. Full modal technology layout+ultimate cost efficiency+ARR explosion certainty brought by the upcoming intensive implementation of multimodal/coding new models, making it have the potential to outperform top model players in the long term even after short-term fluctuations Enterprise capability evaluation: In the competitive analysis matrix (Exhibit 11), Goldman Sachs rated MiniMax's various abilities as follows (out of 5 points): -Organizational efficiency: 5 points (tied with Alibaba for the highest) -Funding: 3 points (weaker than Alibaba's 4 points and Tencent's 5 points) -Training Investments/Model Intelligence: 4 points -Cost efficiency: 5 points (highest score) -Multimodality: 4 points Goldman Sachs acknowledges MiniMax's high organizational and cost efficiency, strong model intelligence, and multimodal capabilities, but points out that its financial strength is relatively weak. Competitive advantage: -Differentiation strategy: model efficiency/cost, agentic capabilities, multimodal/video, global layout -Unique architecture: focusing on data quality, pre/post training, computational efficiency -The full modal capability enables it to maintain cost efficiency and potential gross profit margin higher than its peers in each model iteration (its API revenue gross profit margin continues to be the highest among its peers), which is considered the most important indicator for long-term sustainable business models and expanding the global potential market (TAM) -The parameter scale of minimax M3 is less than 500 billion, with activation parameters of 2.3 billion, which is only one-third of other manufacturers. The DeepSeekV4Pro parameter is 1.6 trillion, and the difference in coding between the two is not significant. If the parameter scale is increased in the future, the intelligence level may not be inferior to other models Risk: -Intense price competition: The price range of minimax-M3 falls perfectly within the most fiercely competitive price range, DeepSeek、 Tencent, Xiaomi MiMo, and others have gained a large number of token shares after lowering prices, and MiniMax's pricing strategy falls into this range -Risk of unlocking in the next six months