Classical infiltration models have been modified to predict infiltration under rainfall sequences. However, a systematic and comparative evaluation of their performance has not been conducted. In this study, the modified Holtan model (MHL), modified Green-Ampt model (MGA), and modified Horton model (MHR) were evaluated on small field plots using sequences of rainfall events produced by a field rainfall simulator. Results showed that both MHL and MGA satisfactorily described the infiltration processes during all rainfall pulses, whereas MHR produced significant errors by overestimating soil drainage rates and thus the recovery of infiltration capacity during rainfall hiatuses. Model MHR was improved by introducing the constraint that drainage occurs only when soil moisture exceeds the value of water content at field capacity and a reduction coefficient to its drainage formula. The improved MHR greatly increased the accuracy of the simulations and performed in a way comparable to the MHL and MGA models. Correlation analysis indicated that the performances of these models did not vary significantly with soil type or topography conditions but was largely affected by the presence of vegetation and the accuracy of their input parameters. Model sensitivity analysis further showed that modeling results were more sensitive to parameters influencing the entire infiltration process than those that only had significant impacts at the initial stage. Therefore, MHL, MGA, and improved MHR are well suited to be incorporated in hydrologic models where an extension to complex rainfall conditions and long-term rainfall simulations is required.