Recently, world endeavors are focused on promoting energy savings by operating both sources and loads at their maximum efficiency points. Thus, this paper presents a novel attempt to optimally determine the operating parameters of an isolated system comprising the proton exchange membrane fuel cells (PEMFCs) stack serving a variable load. A fitness function is adapted to maximize the PEMFCs stack’s efficiency using tuna swarm algorithm (TSA), subjected to set of inequality constraints. A well-known commercial type of PEMFCs stack namely Nedstack PS6 6 kW, is carefully studied over two TSA-based optimization scenarios. The first scenario aims at optimizing five operating parameters, while only two operating parameters are optimized in the second one. Numerical comparisons among the two scenarios are made. It’s worth indicating that the maximum absolute efficiency deviation between both scenarios is equal to 0.8064 at 100 ◦C. Moreover, statistical tests are executed to appraise the performance of the TSA and others. At later stage, the TSA-based results are employed to train and learn an adaptive neuro-fuzzy controller for extracting the optimal operating parameters over wider range of loading conditions, while keeping the goal of maximum efficiency point in order. This allows predicting the optimal values of the operating parameters according to a certain load with a very low time burden, making it able to simulate the real-time load variations effectively and accurately. It can be reported here at low loading values as actual results for example, at 30 % loading condition, the stack’s efficiency is improved from 16.27 % to 63.47 % at 60 ◦C, from 17.24 % to 64.24 % at 80 ◦C and from 18.22 % to 65.26 % at 100 ◦C. While, at load power of 40 %, the FC’s efficiency is enhanced from 21.65 % to 62.72 % at 60 ◦C, from 22.95 % to 63.60 % at 80 ◦C and from 24.25 % to 64.76 % at 100 ◦C. It may be established that via this proposed synergy between TSA and neurofuzzy controller, the efficiency of PEMFCs can be maximized.