OPAL (Object Oriented Parallel Accelerator Library) is a C++ based massively parallel open-source program for tracking charged particles in large scale accelerator structures and beam lines, including 3D space charge, collisions, particle-matter-gas interaction, and 3D undulator radiation. The meticulous parallel architecture allows large and difficult problems, including one-to-one simulations with high resolution and no macro particles, to be tackled in a reasonable amount of time. The current code state as well as the most recent physics advancements and upgrades are discussed, including the unique feature of a sampler for creating massive, labeled data sets with tens of thousands of cores for machine learning. We also demonstrate scalability of our core algorithms up to 4600 GPUs and 32'000 CPUs, as part of our effort to make OPAL exascale ready.