Spontaneous imbibition in porous media plays a crucial role in many engineering applications such as enhanced oil recovery and geological carbon dioxide storage, in which predicting the imbibition rate is of great importance. In this work, we contribute to developing a novel dynamic pore-network model for spontaneous imbibition in porous media. Multiform idealized pore elements have been used to represent complex pore spaces so that our model bears the potential to quantitatively predict spontaneous imbibition for a ‘real’ porous medium. A number of case studies have been conducted to test the pore-network model. We show that the capillary force calculated by the Young–Laplace equation is much smaller than that predicted by the pore-network model. The former is usually used to estimate the capillary force at the wetting front in the widely used single-phase imbibition model. We conclude that using the capillary force calculated by the Young–Laplace equation with the mean pore radius notably overestimates the imbibition rate. Moreover, we verify that a sharp wetting front across a few pores is maintained throughout the primary imbibition process, which is in consistence with experimental observations.