Selective Laser Melting (SLM), has revolutionized the production of complex, high- performance components across industries such as aerospace, automotive, and biomedical. However, the process’s inherent thermal gradients, process uncertainty, and rapid solidification dynamics often lead to defects like porosity and distortions, which compromise part quality. Thus, high-fidelity thermal simulations are critical for predicting these phenomena and optimizing process parameters. Nevertheless traditional numerical methods, such as finite element analysis (FEA), struggle to balance computational efficiency with the resolution required to track the laser spot and the dynamic of the heat-affected-zone. This work presents a novel framework for thermal simulation of the SLM process using the Discontinuous Galerkin method (DGM), accelerated on Graphics Processing Units (GPUs). The DG method’s inherent compact stencil enables precise modeling of transient thermal fields, while its discontinuous basis functions efficiently capture steep thermal gradients near the laser-material interaction zone and allows massively parallel computations due to its weak coupling at the element interfaces. Leveraging GPU parallelism, the framework achieves significant acceleration, enabling high-resolution simulations of large-scale AM builds without compromising temporal or spatial accuracy. The proposed model incorporates temperature-dependent material properties, phase change, and moving laser heat sources to replicate multi-physics interactions inherent to SLM. The DG framework solves the enthalpy equations on a cartesian voxel-grid to further leverage parallelism on GPU architecture and reduce memory footprint. Comparative benchmarks reveals significant speedup over conventional CPU-based solvers, underscoring the method’s scalability for industrial applications.