A Monte Carlo computer code MCNP4A simulation of the TRIGA Mark II benchmark experiment performed in 1992 is presented. It may be noted that this benchmark experiment is one of very few high-enrichment benchmarks available. To minimize errors due to an inexact geometry model, the TRIGA Mark II reactor core was very thoroughly modeled. All fresh fuel and control elements as well as the vicinity of the core were precisely described. MCNP4A input was prepared in such a way that any desired core configuration could be simulated easily. Continuous energy cross-section data from ENDF / B-VI and ENDF / B-V(for nat Cr, natFe, and natNi) libraries and S(α, β) scattering functions from the ENDF / B-IV library were used in our calculations. The differences between ENDF / B-VI and ENDF / B-V evaluations were examined on critical experiments. Most of the steady-state operation experiments were simulated, including two critical experiments, namely, measurements of the excess reactivity of the core, and the determination of control rod worths and fuel element reactivity worth distribution. Excellent agreement with the experimental results was observed.