The hippocampus plays a key role in cognitive map formation. As an abstract model of the world, it has to balance the demands of accurate, yet general, predictions. Despite a long-lasting interest in hippocampal computations, the specific contributions of distinct hippocampal subregions remain debated. Evidence from rodents and humans suggests that subregion CA3 represents event-specific information, while subregion CA1 encodes more generalized information.The free-energy principle entails that neural networks with more generalized representations operate closer to criticality, where external perturbations take longer to resolve. Using a subsampling-invariant method to estimate the distance to criticality based on population spiking activity, we confirm that CA1 operates closer to criticality compared to CA3. Furthermore, we observe that the distance to criticality is significantly modulated by behavioral state and recording day in a subregion-specific way.Overall, our study provides new insights into the relationship between hippocampal function, criticality, and the free-energy principle. If further validated, these findings may help to refine theories of hippocampal function, allow new insights into less-explored hippocampal subregions, and establish the hippocampus as an ideal testbed for exploring ideas derived from the free-energy principle.