Sample size determination for strong ground motion inputs is one of the most critical issues in carrying out nonlinear dynamic analyses. Most popular studies about ground motion sample size at home and abroad often ignore or barely consider the reliability of the structural seismic response estimation. In this paper, the reliability-consistent sample size determination method is presented to ensure a certain reliability of structural seismic response estimation. First, the statistical characteristics of structural nonlinear dynamic response under large ground motions sample for a typical reinforced concrete frame building are analyzed to discuss the difference between the maximum and mean values of structural responses. Then, based on the hypothesis testing of structural response probability distribution models, a sample size determination method for strong ground motion inputs based on the reliability-consistent method is proposed. In addition, the impacts of synthesized ground motions and single-period and multi-period spectral-matching methods for generating synthetic strong motions on the sample size demand are discussed to provide a determination method and estimation basis and ensure reliability and allowable error of structural response estimation using small sample size strong ground motions. The results show that distribution models of structural response highly accept the logarithmic extremum distribution and lognormal distribution hypotheses and completely reject the normal distribution hypothesis. This method is suitable for various types of building structures and ground motion intensities in determining ground motion sample size demand, which is very significant as regards seismic performance and the study of seismic design code.