Abstract: Granular and gas-solid flows are commonly encountered in a range of chemical engineering processes. However, due to the high computational costs, it remains challenging to investigate particle behavior and gas-solid flow hydrodynamics through discrete element simulations. This work developed a graphics processing unit (GPU)-accelerated discrete element method (DEM) that employs an efficient particle collision parallel algorithm and takes full advantage of the parallel structure of GPUs. The DEM code was further coupled with a computational fluid dynamics (CFD) solver through message passing interface (MPI), making it possible to simulate dense gas-solid two-phase flow. The integrated model is verified through three base cases, i.e., a single particle falling and colliding with the wall, two stacked particles compressed between two fixed walls, and a single particle settling in the fluid. The simulation results are in good agreement with the analytic results, indicating the accuracy of the current model. Additionally, this model can accurately predict the particle vertical velocity in a small-scale bubbling fluidized bed and a fully three-dimensional (3D) spout-fluidized bed, confirming its reliability in simulating dense gas-solid flow systems. Furthermore, the GPU-accelerated particle collision parallel algorithm significantly reduces the calculation time and shows great speed-up performance and stability.