In this paper, a model based on genetic programming is presented to credit risk assessment in commercial banks. Empirical results show that genetic programming model is more advantageous than traditional statistical models, neural network models and decision tree models in prediction accuracy and robustness.