Climate change and limited power supplies receive significant incentives to develop alternative paving materials and technologies. Reclaimed asphalt pavement (RAP) and Warm Mix Asphalt (WMA) are technologies that can provide significant benefits to both the environment and the economy. This study used response surface methodology (RSM) to analyze the rutting resistance and moisture susceptibility of asphalt mixtures containing different amounts of RAP and doses and types of WMA. The experimental design was established utilizing the RSM with a central composite design (CCD) for varying RAP dosages (25–50%), WMA amounts (1.5-4%), and WMA types (waxy organic Asphaltan type A ® , and waxy organic Asphaltan type B ® ). The moisture sensitivity and rutting resistance of asphalt samples were evaluated using the Modified Lottman method (AASHTO T 283) and wheel tracking test, respectively. RSM’s statistical and mathematical models were employed to estimate the optimal value for RAP dose and WMA content and type. The results demonstrated that adding RAP to WMA mixtures increased the rutting and moisture resistance of asphalt samples. Also, the analysis of variance (ANOVA) results indicated that the increase in the WMA content led to a significant decrease in the rutting resistance, while the rise in the RAP content contributed to a significant enhancement in the rutting performance of the samples. The statistical outcome also showed that the moisture susceptibility of the mixture decreased significantly after increasing the RAP content, while the increase in WMA content did not have a significant influence on the moisture performance of the mix regardless of the WMA type. The research indicated that rutting resistance and moisture susceptibility have significant correlation coefficients (R2) of > 0.92, indicating that the model is highly correlated with the experimental results. Multi-objective numerical optimization led to the optimal design with 50% RAP and 1.5% WMA-type A. Validation findings indicate strong agreement and model effectiveness, with an error variance of less than 5% for all responses.