With the trend towards improving the building performance, providing realistic occupancy profiles is a key factor in reducing the discrepancies between the actual and simulated energy consumption. Yet, it seems that most of the efforts in modeling occupancy for building performance simulation, disregard the implication of selecting different presence modeling approaches on building performance simulation results. More precisely, the building thermal performance simulation, from the level of workspaces to the building level, and from an hourly peak load to an annual demand calculation may be influenced differently from the occupancy models. If so, selecting among different types of occupancy models (such as the so-called deterministic and stochastic models) can be based on the building performance aspects, which are to be studied. Given this background, this thesis studies the thermal performance of two office buildings (one virtual prototype office building & one real office building), adopting different approaches in modeling occupancy. The simulations are performed regarding Vienna¿s climate and by using EnergyPlus as the simulation software, which enables the incorporations of multiple fixed and randomly generated occupancy profiles into the building models. In order to explore the influence of occupancy modeling assumption on building thermal performance, different building performance indicators for heating and cooling seasons with different reporting frequencies (such as hourly, monthly and annual) were studied. From the results, it is concluded that even though the stochastic presence models offer more realistic distribution of occupancy, there is not a noticeable difference between conventional and stochastic occupants¿ presence models in view of the computed values of annual and peak heating and cooling demands, even by applying different levels of occupant¿s interactions.