The paper investigates real-time output gap estimates for the euro artea obtained from various unobserved components (UOC) models. Based on a state space modelling framework, three criteria are used to evaluate real-time estimates, I.e. standard errors, unbiasedness and conditional inflation forecasts. Real time estimates from univariate moving average filters and from bivariate UOC models based on output and inflation are found to be rather uninformative. Extended models, which employ the information from cyclical indicators and factor inputs, however, improve substantially upon the former models in all criteria. The pessimism on the reliability of real-time output gap estimates expressed in earlier literature may therefore be overstated.