Abstract: This study evaluates the capability of the regional Climate-Weather Research and Forecasting model (CWRF) in simulating the spatiotemporal variations of daily extreme temperature indices from 1980 to 2015 over eight key regions in China based on the homogenization temperature dataset (CN05.1) to provide a scientific basis for improving the model in predicting regional extreme temperatures in China. In this study, we focus on the four percentile-based threshold indices (TX90, TX10, TN90, and TN10) and two duration indices (warm spell duration indicator, WSDI and cold spell duration indicator, CSDI) defined by an expert team on climate change detection and indices. From observation, the annual mean distributions of extreme indices show distinct regional features, and extreme warm events persist longer than extreme cold events. Both warm and cold indices have larger interannual variability in the north, but the warm index shows a warmer trend in most parts of China and the cold index shows a colder trend in North China, where has a more pronounced changing trend of warm and cold nights. The CWRF generally reproduces the observed annual mean distributions of these extreme indices, especially exhibiting superiority in simulating extreme events duration indices, and well simulates the interannual variability and changing trends of extreme indices in most regions. However, several regional biases still exist. For instance, CWRF underestimates the intensity of extremely warm days and cold nights but overestimates WSDI in Central East China and CSDI in Northwest China. In the Qinghai–Tibet Plateau, CWRF tends to underestimate both indices but overestimates the cooling and warming trends of warm days and cold nights, respectively. In East China, CWRF underestimates the cooling trend of cold nights but overestimates the warming trend of warm days and nights.