Solubility of solid compounds is one of the most widely used physicochemical properties in chemical engineering design and experiments. Experimental works for solubility are not always possible because of the small amount of yield available in the phytochemicals extraction. Thus, one interesting perspective is the used of thermodynamic models, which are usually employed for predicting the activity coefficients in the case of solid–liquid equilibria (SLE). Phytochemical compound used in this study is caffeic acid where a comparative study of the MPP-UNIFAC and Pharma Modified UNIFAC were used to predict the solubilities of this phytochemical. The performances of these two activity coefficient models were compared using the experimental solubilities data obtained from the literature in the temperature range of 288 to 323 K and were evaluated by analysing the absolute relative errors (ARE) between the experimental and the predicted values. Moreover, the model errors were also discussed according to the functional groups of the molecules and water as the solvent. In general, the MPP UNIFAC showed better accuracy as compared to the Pharma Modified UNIFAC in predicting the solubility of caffeic acid in water. Nevertheless, both models give very poor qualitative predictions.