When predicting text entry performance time on mobile phones, Fitts’ law has been used to measure finger movement times on the keypad. However, mental activities, such as remembering numbers and visual searches for the right key on keypads, are necessary for text message exchanges. In order to include physical and mental activities to better materialize phone-operating behaviours, another predictive engineering model, the keystroke-level model (KLM), was used with the concept of GOMS. The traditional KLM has six operators but only two operators, K and M, were used for the evaluation of different text entry interfaces in this study. The time for keystrokes ( t K) was calculated with Fitts’-law-based model. To calculate the time for mental activity, phone-operating behaviours were analysed and three mental activities were found: recognizing the intended syllable, recalling the method for composing graphemes and confirming the intended letter. Therefore, the time for accommodating these mental activities is calculated to be 570 ms, which is the sum of time for recognizing and recalling (340 ms) and the time for confirming (230 ms). The KLM-GOMS was developed, validated and utilized to investigate the availability in evaluating different interface alternatives in the design stage and in the comparative study of existing Korean text entry interfaces. The results showed that the model well represented the Korean phone-operating behaviours with the significant statistical test of the average time difference of 6.6%. The results also showed that model was capable of evaluating different text entry interfaces so that it could be used to determine the best Korean text entry interface for the layout standardization of Korean alphabet. In other words, this model could be said to be an efficient tool for the text entry user interface evaluation in the domain of mobile phones without further empirical validations.