Reconstructing particles properties from raw signals measured in particle physics detectors is a challenging task due to the complex shapes of the showers, variety in density and sparsity. Classical particle reconstruction algorithms in current detectors use a multi-step pipeline, but the increase in data complexity of future detectors will reduce their performance. We consider a geometric graph representation due to the sparsity and difference in density of particle showers. We introduce a dataset for particle level reconstruction at the Future Circular Collider and benchmark the performance of state-of-the-art GNN architectures on this dataset. We show that our pipeline performs with high efficiency and response and discuss how this type of data can further drive the development of novel geometric GNN approaches.