Several industrial contexts require software engineering methods and tools able to handle large -size artifacts. The central idea of abstraction makes model-driven engineering (MDE) a promising approach in such contexts, but current tools do not scale to very large models (VLMs): already the task of storing and accessing VLMs from a persisting support is currently ine cient. In this paper we propose a scalable persistence layer for the de-facto standard MDE framework EMF. The layer exploits the e ciency of graph databases in storing and accessing graph structures, as EMF models are. A preliminary experimentation shows that typical queries in reverse-engineering EMF models have good performance on such persistence layer, compared to le-based backends.