In this article we analyze the accuracy and stability of short-run inflation forecasts for Chile coming from Extended Seasonal Arima (ESARIMA) models. We compare ESARIMA forecasts to those coming from surveys and traditional time series benchmarks available in the literature. Our results show that ESARIMA based forecasts display lower out-of-sample Mean Squared Prediction Error than forecasts coming from traditional benchmarks when the predictive horizon ranges from 1 to 4 months. At longer horizons, the worst models from the ESARIMA family are outperformed by the best univariate traditional benchmarks. We obtain opposite results when comparing ESARIMA outcomes to survey-based forecasts: the survey provides more accurate forecasts at every single horizon. Our results are, in general, statistically significant at usual confidence levels. We also notice that ESARIMA forecasts are more stable than traditional time series methods but less stable than survey-based forecasts.