Zhipu's GLM-5 tops open-source AI rankings
Zhipu AI, listed in Hong Kong as Knowledge Atlas Technology JSC, recently released the latest iteration of its large language model, GLM-5. According to a Jefferies research note cited in financial press coverage, GLM-5 surpassed a rival offering from Moonshot AI released just weeks earlier to claim the top spot among open-source models worldwide on benchmarking site Artificial Analysis. The Jefferies note described the ranking as the highest ever achieved by a Chinese AI laboratory on the platform.
Hong Kong debut draws investor enthusiasm
Zhipu and MiniMax Group both debuted on the Hong Kong stock exchange in January and have emerged as the most prominent Chinese pure-play AI stocks for investors. Zhipu's shares have climbed 524% since listing, while MiniMax has risen 488%, reflecting a dearth of listed global large language model developers. Bullish coverage from Wall Street banks, including Morgan Stanley, has reinforced the rally, even as Chinese investors treat each new AI release as a catalyst not only for developers but also for downstream users of new tools.
Zhipu model aids US cybersecurity response
Hugging Face, the New York-based open-source AI community, disclosed that it turned to Zhipu AI's GLM-5.2 model to analyze data from a cybersecurity incident involving an autonomous malicious agent built with OpenAI technology. According to a Reuters report, several leading US AI models declined to assist with the incident response because their safety filters could not distinguish between defenders and attackers. A source familiar with the matter told the Global Times that the case illustrates how enterprises are broadening their evaluation of foundation models beyond raw capability to include production deployment, engineering usability and complex task handling.
Divergence from global AI scare trade
While US markets have sold software companies and wealth managers on fears that rapid AI advances will erode incumbent business models, Chinese investors have largely shrugged off the so-called "AI scare trade." Saxo Markets chief investment strategist Charu Chanana said China's market remains focused on what AI can aid rather than what it can take away from incumbents, while Allspring Global Investments portfolio manager Gary Tan pointed to China's relatively insulated competitive landscape, where regulatory constraints and geopolitical tensions limit foreign participation by AI-related companies.
Chinese open-source models gain global traction
Data from Hugging Face show that Chinese models now account for roughly 41% of downloads on the platform, surpassing US-origin models in both monthly and total downloads. The latest OpenRouter ranking also reflects growing adoption of Chinese open-source large language models. The trend has prompted some Chinese analysts, including Chen Jing, vice-president of the Technology and Strategy Research Institute, to argue that the closed approach favoured by US labs imposes safety filters that can block defenders from analysing real attack data, while open models give security teams more flexibility.
Share this article
