Radio frequency (RF) cable assemblies play a crucial role in high-frequency communication systems, and their passive intermodulation (PIM) performance is of paramount importance. In this article, a novel PIM model containing distributed and point-source nonlinearities is provided to investigate the effect of coaxial cable length and connectors on the intermodulation powers of cable assemblies. Based on the structure of the cable assembly, a point-source PIM model in connectors and a distributed-nonlinear model of RF coaxial cables are proposed, respectively. Considering the phase shift and attenuation induced by the electrical length, an RF cable assembly PIM model, which can predict the reverse third-order intermodulation power for different cable lengths is established. A series of equivalent circuit models for point-source, distributed and comprehensive nonlinearities are presented to simulate the variation of PIM behaviors with distributed structures and point source power. In addition, a method for calculating the coefficients of nonlinear polynomial model of coaxial cable is proposed. RF cable assemblies with different types of connectors on both ends and varying lengths are designed and produced as the samples for experimental verification. The measured results are consistent with the theoretical prediction. For different types of coaxial connector samples, the third-order intermodulation average error is 0.71 dB. Both simulation and experimental results show that changes in cable length causes variations in assembly intermodulation powers, with a maximum fluctuation exceeding 10 dB.