Path loss (PL) models are critical mathematical frameworks for characterizing signal attenuation in wireless communication systems, and the environments, distance, frequency, and physical obstructions usually influence it. Accurate PL prediction is vital for optimizing the design, deployment, and operation of next-generation networks, especially at millimeter-wave (mmWave) frequencies, which is a key enabler for 5G communication networks. Despite prior advancements in developing and validating 5G PL models, limited research exists on their empirical evaluation with region-specific radio propagation data, especially for the Northern part of Nigeria. This study thus addresses this by identifying the most suitable PL model for 5G communication systems at the mmWave frequencies in Zaria, Nigeria. Using empirical data collected at 28 GHz, 36 GHz, and 48 GHz, the performance of five prominent PL models: Close-In (CI), Alpha-Beta-Gamma (ABG), Stanford University Interim (SUI), Free-Space Path Loss (FSPL), and the Floating Intercept (FI) models was evaluated. Results indicate that the CI model offers superior accuracy across the measured frequency bands with an average Root Mean Square Error (RMSE) of 4.66 dB across the considered frequencies. Following closely to the CI model is the ABG model, which has an average RMSE of 7.61 dB. However, the SUI, FI, and FSPL models showed a very poor estimation of the Path Loss, with an average RMSE of 30 dB. These findings provide critical insights for network engineers, enabling the development of precise and efficient PL estimations to enhance the planning, deployment, and optimization of 5G mmWave networks in urban environments in Nigeria and related environments.