The development of the three major model companies in China is completely different from the image they have started to give to everyone The most academic (Zhipu) → the strongest revenue MiniMax, who understands commercialization the most → has the weakest revenue Kimi, the most popular among investors, returns to academia in the end GLM (Zhipu AI) - Born in an ivory tower, but the most profitable The most academic team, but also the earliest and highest revenue team in commercialization -Integrated: Tang Jie, Zhang Peng, Liu Debing, pure academic lineage of Tsinghua KEG Laboratory -Two wings: Introduce Wu Weijie (formerly known as Feishu/GE/Cisco), Zhang Fan and other AWS/Amazon commercialization fierce generals to be responsible for sales -Coming from the most academic background, but with a revenue of 724 million yuan in 2025, the strongest among the three, and an open-source model ranking among the top three in the world Keywords: academic background+sales iron army, the most underestimated combination, current king MiniMax - Experienced IPO, but revenue is the most embarrassing It turns out that SenseTime was afraid of doing B2B business, and the first to flourish in entrepreneurship was in segment C. The team members also have financial backgrounds in corporate mergers and acquisitions, IPOs, etc. -Yan Junjie comes from SenseTime (Vice President), and the core team consists of all SenseTime members: Zhou Yucong (SenseTime Algorithm), Yun Yeyi (SenseTime Business) -Multimodal is the most aggressive (with five parallel lines), with Talkie covering 212 million users overseas -I am the most knowledgeable about commercial monetization, but my revenue is actually the worst among the three (with a cumulative loss of $1.32 billion over three and a half years) Keywords: Industry practicality, multimodal, multi line product development, once a pride but now a burden, cost-effectiveness model not outstanding enough Kimi (Dark Side of the Moon) - the purest academic founding team After making radical attempts at commercialization and realizing that model capability was the most crucial, the team turned to relentlessly pursuing models -Founder Yang Zhilin, with a bachelor's degree in computer science from Tsinghua University, studied under Ruslan Salakhutdinov, the head of Apple AI, and William Cohen, the chief scientist of Google; Write two key papers as the first author, Transformer XL and XLNet, which laid the foundation for the era of pre training -4/5 of the team comes from Tsinghua University and CMU, with members covering top global labs such as FAIR and Google AI -In 2024, the investment flow was the strongest (nearly 200 million in a single month), and after Zhang Yutong took office, he called for a halt and returned to the technology premium route Keywords: academic geek, returning to its true colors after a brief commercialization