Predicting asphalt pavement performance is an important matter which can save cost and energy. To ensure an accurate estimation of performance of the mixtures, new soft computing techniques can be used. In this study, in order to estimate the stiffness property of Polyethylene Terephthalate (PET) modified asphalt mixture, different soft computing methods were developed, namely: support vector machine-firefly algorithm (SVM-FFA), genetic programming (GP), artificial neural network (ANN) and support vector machine. The Support Vector Machine-Firefly algorithm (SVM-FFA) is a metaheuristic search algorithm developed according to the socially dashing manners of fireflies in nature. To develop the models, experiments were performed. The process, which simulates the mixtures' stiffness, was created with a soft computing method, the inputs being PET percentages, stress levels and environmental temperatures. The performance of the proposed system was confirmed by the simulation results. Soft computing methodologies show very good learning and prediction capabilities and the results obtained in this study indicate that the SVM-FFA contributed the most significant effect on stiffness performance estimation since the SVM-FFA model had a better correlation coefficient than the SVM, ANN and GP approaches. R 2 and RMSE were utilized for making comparisons between the expected and actual values of SVM-FFA, GP, ANN and SVM. The proposed SVM-FFA methodology predicted the output values with 254.4743 (mm/day) and 0.9957 RMSE and R 2 respectively.