We present a fast wavelet-based time-domain modeling technique to study the effect of electromagnetic (EM)-wave propagation on the performance of high-power and high-frequency multifinger transistors. The proposed approach solves the active device model that combines the transport physics, and Maxwell's equations on nonuniform self-adaptive grids, obtained by applying wavelet transforms followed by hard thresholding. This allows forming fine and coarse grids in the locations where variable solutions change rapidly and slowly, respectively. A CPU time reduction of 75% is achieved compared to a uniform-grid case, while maintaining the same degree of accuracy. After validation, the potential of the developed technique is demonstrated by EM-physical modeling of multifinger transistors. Different numerical examples are presented, showing that accurate modeling of high-frequency devices should incorporate the effect of EM-wave propagation and electron-wave interactions within and around the device. Moreover, high-frequency advantages of multifinger transistors over single-finger transistors are underlined through numerical examples. To our knowledge, this is the first time in the literature a fully numerical EM-physics-based simulator for accurate modeling of high-frequency multifinger transistors is introduced and implemented.