AbstractMolecular dynamics (MD) simulations of creep generally face the problem that the creep most often evolves on time scales hard to capture with MD due to their typically short time step size. Consequently, MD studies of creep often use unrealistically high temperatures and stresses and simplified atomistic models to make creep-like processes happen on computationally accessible time scales. Apparently, this compromises the physical reliability of such studies. To alleviate this problem, we designed an MD model of titanium aluminide (TiAl) with a microstructure matching at least many of the key parameters of experimentally observed microstructures. We applied this MD model with stresses much lower than the ones used in most previous creep studies (well below yield stress) and in the temperature range , with melting temperature . Compared to typical previous MD studies, this much more realistic setup produces creep rates more than three orders of magnitude smaller and thus much closer to reality. We identified the driving mechanisms of primary creep on the nanosecond scale that agree very well with recent experimental observations, thus contributing towards the overarching goal of bridging the gap between atomistic creep simulations and continuum-scale creep simulations for engineering applications.