This manuscript presents a novel application of the recently developed Puma Optimizer (PO) for identifying the unknown parameters in Mann’s model, which is widely used for characterizing the behavior of Polymer Electrolyte Membrane Fuel Cells (PEMFCs). The proposed PO-based methodology is rigorously evaluated using three test cases. One test case involves experimental I–V measurements under various operating conditions from a commercial PEMFC stack, the Horizon H-100 (100 W), which was assembled and tested in the laboratory. The other two cases are established benchmark PEMFC systems, the Ballard Mark V 5 kW and BCS 500 W units. Comprehensive statistical analyses over multiple independent runs are performed to confirm the consistency and reliability of the PO-based approach. To evaluate the optimizer’s accuracy and robustness, comparisons are made with three other metaheuristic algorithms: two recently developed methods, the Propagation Search Algorithm (PSA) and the Walrus Optimization Algorithm (WOA), and a widely recognized one, the Slime Mold Optimizer (SMO). The comparisons show that the PO optimizer consistently achieved the lowest total square error (TSE) across all test cases, outperforming the other algorithms. Specifically, it develops the lowest values of 0.835811 for the Ballard Mark V 5 kW, 0.011281 for the BCS 0.5 kW, 0.711452 at 30 °C, 0.886144 at 35 °C, and 2.057015 at 40 °C for the Horizon H-100 fuel cell. These findings confirm that the PO is a reliable and effective tool for parameter estimation in PEMFC modeling under both simulated and real-world experimental conditions.