This study aims to compare the performance of Self-Exciting Threshold Autoregressive (SETAR) and Markov Switching Autoregressive (MSAR) models in modeling and predicting inflation data in Indonesia. Inflation is one of the important economic indicators that requires accurate forecasting methods to make the right policy decisions. In this study, both models are applied to monthly inflation data in Indonesia. The SETAR model is a nonlinear autoregressive model that takes into account regime changes in the inflation rate based on a certain threshold value. Meanwhile, the MSAR model assumes that inflation data can move between regimes with a certain probability governed by a Markov process. Both models are evaluated based on the Akaike Information Criterion (AIC) and Mean Absolute Percentage Error (MAPE) values, with the aim of finding the model that provides the best forecasting results. The results show that the SETAR model has a lower AIC value of -1658 with a MAPE of 18.19% compared to the MSAR model of -832.6177 and MAPE of 25.46%, which indicates that SETAR is superior in modeling and predicting inflation data in Indonesia. This finding makes a significant contribution in choosing a more accurate forecasting method for inflation data and can help in planning better economic policies.