Mould growth results from a complex interaction between environmental factors, material properties, and mould fungi characteristics. These interactions must be considered during the design, construction and maintenance of a building to prevent growth. Mould prediction models aim to predict whether mould will grow on a specific material in a part of building with a known, or simulated, relative humidity and temperature. They are often used in the design phase. Several models are available. There is limited research on the performance of the models in real buildings. This study aimed to evaluate six different models, using data from five building parts. The predictions on whether mould growth was expected or not were compared to actual mould growth observations on five building materials. The study was performed as a round-robin. Most models underestimated the possibility for mould when humidity and temperature varied a lot by time. The outcome also depended on the end-user, who needs to make assumptions and parameter values choices on, for example, material susceptibility for mould growth. Therefore, using the same climate data, mould growth prediction may differ depending on who makes the prediction. One model, MOGLI model, where input data comes from laboratory tests and no such assumptions must be made, predicted correct in most cases. One conclusion of the study is that when predictions are made in practice, the results must be used cautiously. More knowledge is needed to understand, and more accurately model, the relationships between the moisture and temperature variations in buildings and the risk for mould growth. •Results from mould prediction models were compared to results from field tests.•Choices of input parameters varied among users, and this had a significant impact on results.•Dynamic mould prediction models often underestimated the possibility of mould growth.•The PJ-model was shown to give the best predictions in comparison with field results.