The aims of this work were modelling the effect of water stress intensity during post-veraison on table grape quality and yield, as well as predicting berry quality at harvest using a machine learning algorithm. The dataset was obtained by applying different irrigation regimes in two commercial table grape vineyards during seven growing seasons. From these data, it was possible to train and validate the predictive models over a wide range of values for the independent (water stress intensity and fruit load) and dependent (firmness and berry color) variables. The supervised learning algorithm Gaussian Process Regression allowed us to predict the variables with high accuracy. It was also determined that a reduction in irrigation of up to 40% during post-veraison, compared to vines without water limitations, and the accumulation of the water stress integral of up to 30 MPa per day, linearly increase the irrigation water use efficiency (IWUE) and promote higher berry color and firmness. The severe water scarcity and the increasing uncertainty about the irrigation water availability for the season that farmers are facing highlight the advantage of incorporating these validated techniques into agricultural decision making, as they allow for the planning of cultural practices and criteria to increase the IWUE and crop sustainability.