Model-based testing (MBT) uses behavioral models of the system to derive tests cases. A typical approach is to generate the labelled transition system (LTS) of a model and then using model checking to discover paths satisfying certain test requirements. A significant challenge is state space explosion and the complexity of model checking. Some authors have proposed meta-heuristic search-based approaches to cope with this problem, exploring only a small portion of the LTS to produce paths that cover a maximum of test objectives. Although yielding acceptable results on small case studies, these approaches don’t scale well. MBT using graph transformation can use both the behavioral information and the graph structure of states and rules to de-fine and evaluate test objectives, but approaches have the same limitations as LTS-based MBT in general, often exacerbated by the more complex nature of graph-based LTS. In this paper, we introduce a method based on deep reinforcement learning to generate test suites for systems specified through graph transformations. We use the reward/penalty mechanism of the reinforcement learning to op-timize the selection of moves within the state space, enabling the generation of test cases based on prior decisions. Our goal is to achieve greater coverage of test objectives while minimizing the size of the test cases. The method has been implemented in GROOVE, an open-source toolset for designing and model checking graph transformation systems. Experimental results on well-known case studies demonstrate that our approach generates test cases with improved coverage scores while requiring less cost.