Driven by the growth of variable renewable electricity generation, the increasing electrification of heating and cooling of buildings holds a lot of demand-side management potential worldwide. However, building-stock-scale models depicting this flexibility for large-scale energy system optimisation models are lacking in literature. Thus, we've been developing an open-source tool called ArchetypeBuildingModel.jl specifically for this purpose, aggregating building-stock-level statistical and technical data into desired simplified lumped-capacitance thermal models compatible with existing open-source energy system modelling frameworks. This paper aims to demonstrate the feasibility of such simplified thermal models via comparisons against dedicated building simulation software, as well as examine the sensitivities arising from the key modelling and parameter assumptions. In conclusion, parameter and modelling assumptions corroborated by existing literature achieved reasonable performance across all tested buildings and nodal configurations, although caution is advised when modelling buildings with structures heterogeneous in terms of thermal mass. Nevertheless, even at its worst, model performance was deemed plausible.