Occupant-controlled window opening plays a significant role in exposure to indoor environmental hazards such as high temperatures and PM2.5. Both the indoor and outdoor environments can affect window opening behaviour of building occupants, while variation in this behaviour is often poorly defined in building simulations. This study examines the robustness of six existing behavioural models of window opening in predicting window-use patterns in Hong Kong flats, and the impact of behavioural models on the accuracy of indoor temperature and indoor PM2.5 concentration prediction. The behavioural models were implemented in the EMS application of EnergyPlus. The predicted window-use patterns were compared to those of filed measurements, whilst the predicted indoor temperatures and indoor PM2.5 concentrations were benchmarked against those predicted by the EnergyPlus model with the actual occupant window opening behaviour. Results indicate that the ability of individual behavioural models to accurately predict occupant window opening behaviour varies depending on modelling approach, outdoor environmental variable and simulation time-step size. The model providing the best estimate of the duration of window-opening is able to best contribute to accurate predictions about indoor temperatures and indoor PM2.5 concentrations. Results also show that a static temperature threshold for window-opening (a common method of modelling occupant window opening behaviour) can lead to a significant difference between the actual and predicted window-use patterns (and therefore indoor temperatures or indoor PM2.5 concentrations), underlining the importance of improving model inputs for occupant window opening behaviour.