Since 2023, Chinese AI models have consistently lagged behind their US counterparts, with an average gap of seven months in capabilities. The Epoch Capabilities Index (ECI) ranks AI models based on performance, and during this period, all models at the cutting edge have emerged from the United States. The gap between the two countries ranges from four months to a maximum of 14 months. This disparity is particularly pronounced between proprietary US models and open-weight Chinese models, as most leading Chinese models fall into the latter category. To measure this gap, researchers analyzed the highest ECI scores for both countries' models upon their release, excluding the initial models that likely did not represent the true frontier. The methodology involved tracking the best-performing Chinese model daily and comparing its ECI score to that of leading US models. The first Chinese model to exceed the performance of OpenAI's GPT-4 appeared in May 2024, marking a 14-month delay. As of now, no Chinese model has surpassed the capabilities of OpenAI's o3 model, which was introduced in April 2025. This ongoing trend raises questions about the competitive landscape in AI development between the US and China.