Colour appearance models (CAM) such as CIECAM97s and CIECAM02 were originally developed to predict the change of colour appearance under different viewing conditions. This paper describes the development of CAM extensions for evaluating colour–differences. The performances of the original CAMs and their extensions were tested using two types of colour-difference data: large and small magnitude colour-differences. The results showed that the CIECAM02 based models in general gave a more accurate prediction than the CIECAM97s based models. The modified version of CIECAM02 gave satisfactory performance in predicting different data sets (better than or equal to the best available uniform colour spaces and colour-difference formulae). This strongly suggests that a universal colour model based upon a colour appearance model can be achieved for all colorimetric applications: colour specification, colour difference and colour appearance.