Computational molecular science is a very computationally intense discipline, and the use of parallel programming and high-performance computers well established as being necessary to support research activities. Recently, graphical processing units (GPUs) have garnered substantial interest as alternative sources of high-performance computing capability. These devices, although capable of very high rates of floating-point arithmetic, are also intrinsically highly parallel processors and their effective exploitation typically requires extensive software refactoring and development. Here, we review the current landscape of GPU hardware and programming models, and provide a snapshot survey of the current state of computational molecular science codes ported to GPUs to help domain scientists and software developers understand the potential benefits and drawbacks of this new computing architecture. © 2012 John Wiley & Sons, Ltd.