We examine whether seven state-of-the-art European regional air quality models provide daily ensembles of predicted ozone maxima that encompass observations. Using tools borrowed from the evaluation of ensemble weather forecasting, we analyze statistics of simulated ensembles of ozone daily maxima over an entire summer season. Although the model ensemble overestimates ozone, the distribution of simulated concentrations is representative of the uncertainty. The spread of simulations is due to random fluctuations resulting from differences in model formulations and input data, but also to the spread between individual model systematic biases. The ensemble average skill increases as the spread decreases. The skill of the ensemble in giving probabilistic predictions of threshold exceedances is also demonstrated. These results allow for optimism about the ability of this ensemble to simulate the uncertainty of the impact of emission control scenarios.