GDP is one of the most important indices in measuring the socioeconomic development status of a country, and thus, an accurate estimation of GDP is vital in formulating valid development strategy for a country. However, the geographical data used for GDP simulation in various studies are different, and it is not clear which geographical data is more conducive to GDP simulation. In this study, both ordinary least squares regression ( OLS) and geographical weighted regression ( GWR) methods were conducted in order to simulate the GDP of 2 848 counties in China, in which, the performance and modeling capabilities of multi-sourced open data, including nighttime satellite data, points of interest ( POI) data, Tencent‘s social user location data and built up urban area data, were explored and assessed. The experimental results showed the following findings: ① The overall accuracies of the simulated GDP at counties and cities in China could be achieved above 74% and 87% , respectively, when multi sourced open data, including nightlight satellite data, POI data and Tencent’s user data were utilized by GWR method. ② By comparing different indicators, it is found that the point of interest ( POI) performs better than the other indicators in modelling the actual GDP at the county level in China. ③ The results also indicated that Tencent‘s social user location data has high potential in modeling the GDP distribution in the western part of China, which indicated that the modeling accuracy of GDP in China could be further improved using both Tencent‘s social user data and POI data. The results and findings in this study could provide insights in understanding how to better simulate and model GDP distribution in China.