To estimate the energy performance of dwelling envelopes using non-intrusive monitoring methods, disaggregating energy demand profiles is required. In dwellings with instantaneous gas heaters and gas stoves, the demand profiles for space heating, domestic hot water, and cooking should be disaggregated from the total gas demand. While some disaggregation methods exist, it is unclear how accurate they are. This paper contributes to this gap by collecting data on space heating, domestic hot water, and cooking demand profiles from eight Belgian dwellings through a survey and measurements of the total gas demand, water temperatures at the heater, and - for three of them - hot water flows. The total gas demand was disaggregated based on these data and the space heating demand profiles compared with those disaggregated by five models: 1) a survey-based model; 2) the energy signature curve combined with singular spectrum analysis; 3) pattern recognition; 4) a peak filling model; and 5) a grey-box model with a switching model. The models’ disaggregation accuracy was determined for datasets that differ in measurement duration, sampling interval, and measurement season. Generally, the models underestimate the contribution of heating to the total gas demand. For hourly sampling intervals, the survey-based model performs best. For 15 min intervals and winter data, the grey-box with switching model performs better and for these intervals but autumn data, the energy signature curve. For 5 min sampling intervals, pattern recognition performs best.