Cosmic voids have been shown to be an effective probe of cosmology, complementary to galaxy clusters. In this work, we present a simple theoretical framework for predicting of the size of the largest voids expected within a given redshift and volume. Our model is based on the exact extreme-value statistics which has previously been successfully applied to massive galaxy clusters. We implement our formalism using the void-abundance models and compare the extreme-void predictions to simulations and observations. We find that the simplest void models can only explain the extreme-void abundance with ad hoc parameter adjustments. We argue that the prediction extreme-void distribution should be used as an additional test on theories of void abundance.