Goldman Sachs has identified its top Chinese artificial intelligence models, with only one currently available on public markets. The investment bank initiated coverage on Zhipu (also known as Knowledge Atlas), a Hong Kong-listed company, assigning it a price target of 1,880 Hong Kong dollars ($239.83), representing nearly 15% potential upside. Zhipu has experienced significant growth since its listing in January and recently garnered attention for its open-sourced GLM-5.2 model, which is considered a strong contender against Anthropic's Fable 5 on various performance metrics.
"With its latest GLM5.2 model reaching near-frontier performance that has seen significant ramp-up in domestic enterprise & global SME adoption, we believe its extensive usage by coders will enable Zhipu to sustain high frequency of further model upgrades, and thereby solidify its leading position in enterprise/coding in China," stated the analysts.
While Zhipu received a neutral rating, Goldman Sachs' other favored Chinese AI model companies are the privately held Deepseek and ByteDance. These assessments were based on factors including time to market, arena score, valuation, and pricing. The research also highlighted AI video generation capabilities, where ByteDance reportedly excelled.
According to the Goldman report, Zhipu's GLM and DeepSeek's models generally outperformed those from Alibaba, Tencent, and Minimax, particularly in terms of time to market and Arena score. This has been reflected in market performance, with Zhipu shares surging 70% in Hong Kong over the last 60 trading days, while Minimax has plummeted over 70%. Alibaba has declined nearly 10%, and Tencent is down about 5%.
The analysts observed that "China's AI open-source/open-weight models are reaching a critical point of intelligence performance vs. global proprietary models." They further noted that "Agentic AI is driving explosive demand for these value-for-money models at the lower-end. Access to computing will be a swing factor, where US/China regulations, balance sheet and inference efficiency are key."
