MINIMAX融资金额多但亏损更大 多项研发指标低于智谱 语言模型评估排全球第一是否有参考价值？

Core Viewpoint - The domestic large model companies Zhipu and MINIMAX are competing for the title of "first large model stock" in Hong Kong, with Zhipu showing slightly higher revenue while MINIMAX has a larger fundraising amount but greater losses [1][2]. Group 1: Financial Performance - Zhipu's revenue for 2022-2025 is projected to be 0.57 billion, 1.25 billion, 3.12 billion, and 1.91 billion RMB, with net losses of -1.43 billion, -7.88 billion, -29.56 billion, and -23.51 billion RMB, totaling a loss of 62.38 billion RMB over three and a half years [3]. - MINIMAX's revenue for the same period is 0, 0.03 billion, 0.31 billion, and 0.53 billion USD, with net losses of -0.74 billion, -2.69 billion, -4.65 billion, and -5.12 billion USD, totaling approximately 94 billion RMB in losses over three years and three quarters [3][6]. - As of the disclosure date, MINIMAX has raised approximately 109 billion RMB through equity financing, while Zhipu has raised about 83 billion RMB [3][6]. Group 2: Research and Development - MINIMAX's R&D spending from 2022 to 2025 is 0.11 billion, 0.7 billion, 1.89 billion, and 1.8 billion USD, accounting for 0%, 202.2%, 619.1%, and 337.4% of total revenue respectively [7]. - Zhipu's R&D spending for the same period is 0.84 billion, 5.29 billion, 21.95 billion, and 15.95 billion RMB, representing 147.0%, 424.7%, 702.7%, and 835.4% of total revenue respectively [7]. - As of June 30, 2025, Zhipu has 657 R&D personnel, making up 74.4% of total employees, while MINIMAX has 284 R&D personnel, accounting for 73.77% of its total workforce [8]. Group 3: Intellectual Property - MINIMAX has registered 74 patents and 72 copyrights, while Zhipu has 86 patents (84 of which are invention patents) and 160 copyrights, indicating that MINIMAX's intellectual property is significantly lower than Zhipu's [9]. - The lower number of patents and copyrights for MINIMAX does not necessarily indicate inferior product capability compared to Zhipu, but the comparison of data is valuable [9]. Group 4: Model Performance - MINIMAX's language model Speech-02-HD ranks first globally with a score of 1174, surpassing OpenAI and Google [12][13]. - However, there are concerns regarding the credibility of the ranking provided by Artificial Analysis, as it contrasts with general market perceptions and other rankings where MINIMAX does not appear in the top twenty [14][15].