Abstract Within an energy system where nuclear reactors will likely play a critical role, it is fundamental to have access to accurate and reliable computational tools that can predict the plants' behavior under different operating conditions; compared to other energy sources, models and analysis methods for nuclear systems should be able to provide, among other things, detailed information on reactor criticality and fuel evolution. Thanks to the advancements in computational hardware, using three-dimensional codes to obtain a local description of the reactor core has now become feasible, not only for what concerns deterministic codes but also for Monte Carlo (MC) based codes. Those computational methods must be compared with experimental measurements to assess their reliability. For this reason, the 3D Monte Carlo code serpent is currently being validated for light water reactor fuel cycle simulations. This work will compare the isotopic concentrations measured in a Post-Irradiation Experiment (PIE) and the results of the Monte Carlo routine, examining the Takahama-3 fuel assembly test case. Since key information related to the history of the plant under consideration is available, it was possible to follow the depletion cycles of two fuel rods (SF95 and SF97) placed in two separate assemblies. From literature reports, more than 35 nuclide species have been measured at different axial locations by destructive analysis following several radiochemical techniques. A sensitivity analysis, aiming to evaluate the impact of design features on the results, was carried out investigat-ing the cross section libraries, the simulation time Discretization and the imposition of an axial time-varying temperature. During the process, systematic sources of geometry-related errors were analyzed as well. Over-all, the model showed good agreement with the experimental data under an acceptable error threshold. The sensitivity studies also showed how the prediction capability could be increased up to +6%, adopting a realistic temperature mesh for the fuel instead of a uniform temperature approach. Finally, this analysis will focus more deeply on the Antimony-125 evolution, whose concentration is the one with the worst agreement to the experimental dataset not only for this work but also for other benchmarks available in the literature: results will show how the prediction error for this nuclide is systematic for all considered codes, and look for possible causes.