Turbulence enhances mixing drastically. Sometimes, a simple model of that enhancement is all we need for engineering and geophysical applications. We can ignore the small-scale details. Some other times, however, the small-scale details control the large-scale behavior. The multi-scale nature of turbulence unfolds and challenges all efforts to describe turbulence in simple ways. Supercomputing has given us the opportunity to study this multi-scale nature with unprecedented accuracy, and thereby to advance turbulence modeling to what is necessary for the societal demands during the coming decades.