OpenMC has been mostly ported to GPU architectures by the OpenMC team in a fork ( https://github.com/exasmr/openmc ), however there’s a bit of a feature gap between this and OpenMC in the main repository due to some incompatibilities between external libraries (e.g. libMesh for unstructured mesh tallies) and the GPU port. The GPU port required a near-complete rewrite of OpenMC as well due to limitations (at the time) of OpenMP target offloading support, and so merging it into main wasn’t considered until compiler support improved.
The OpenMC team recently received funding to investigate / complete this port. More information can be found here: Status/Plans to merge GPU OpenMP branch