are Chinese ai models better than American?

Chinese AI models are closing the gap with American counterparts and are surpassing them in areas like cost and accessibility, but they generally lag in overall frontier performance. Since 2023, Chinese models have trailed the U.S. frontier by an average of 7 months, though experts predict China could match U.S. capabilities in 2025 [1][5]. Models like DeepSeek's R1 and Z.ai's GLM-5 rival top U.S. offerings from OpenAI and Google, achieving 80-90% of frontier performance at significantly lower training and inference costs—often 1/10th of Western equivalents—while being released as open-source under permissive licenses [6][9]. This has enabled Chinese open models to capture about a third of global AI usage by late 2025, eroding U.S. dominance in closed-source markets and pressuring American firms on pricing [2][7][11]. However, a performance gap persists due to China's constraints on computing power and capital, making it unlikely for China to sustainably surpass the U.S. in the near term [8][12].