Magnetic tweezers are a powerful tool for the study of biomolecules due to their ability to simultaneously measure the characteristics of molecules in parallel and with high accuracy. The magnetic tweezers use video-based particle tracking to determine the three-dimensional position of surface tethered magnetic probe particles and thereby deduce elastic properties of the tether molecules. However, most instruments are limited in their throughput to tracking due to the computational complexity of the 3D position finding algorithms. Here, we combine a GPU and the CUDA architecture with a magnetic tweezers instrument, and demonstrate an initial two-fold improvement in data throughput over conventional CPU-based approach. Out results demonstrate the ability to track nearly 5000 beads per second in 3d. These tests indicate that certain tasks, such as high speed or highly parallel tracking, that were once impossible to accomplish in real time are now possible and available to those who have access to a GPU. Furthermore, the potential to use the GPU for bead tracking is still in its infancy, and we expect to continue to multiply the performance with improvements in the code, and greater utilization of the GPU.