This paper describes the impact of perimeter column boundary conditions in the modelling of multi-storey structures when subject to fire. An idealised model of a building section is created with a range of column boundary conditions; the sensitivity of results are then analysed. It is found that considering only a limited number of floors results in an overestimation of the moment in the perimeter column by up to 50% under standard heating, but has relatively little impact on beam displacements—less than 5% after standard heating. It is found that the proximity of the boundary conditions has less impact on the results as column stiffness reduces, or if moment connections are provided. A range of possible simplified boundary conditions are considered and it is concluded that the cruder restrained simplifications are likely to lead to an overestimation of column moment sometimes in excess of 50%, but have relatively little impact on floor deflection. Of the idealised boundary conditions analysed, it was found that representing 3 floors and 4 heights of column gives the most significant improvement in accuracy in comparison to a single storey model. Models with pinned boundary conditions ±1-floor (relative to the fire floor) and fixed boundary conditions at ±1.5 or ±2-floors (relative to the fire floor) gave a more accurate result than simply fixing at ±1-floor. It was found that when connections between the beam and the column are pinned, the model with boundary conditions pinned at ±1-floor and fixed at ±2-floors gives a more accurate result than the model with fixed conditions at ±1.5-floors. The results of the study are presented to allow modellers to make an informed judgement about the level of complexity with which to represent a structure above and below the fire-heated floor.