Chuanfei Dong, speaking in the Institute for Advanced Study's Astrophysics Seminar, surveys the numerical methods used to model plasmas that behave differently depending on scale. He walks through magnetohydrodynamic, multi-moment multifluid, and particle-in-cell approaches, along with hybrid schemes like MHD-AEPIC that combine fluid and kinetic treatments in the same simulation. Examples include local simulations of magnetic reconnection and turbulence, and global models of star-planet interactions both within the solar system and around other stars. Dong also describes recent work using machine learning to embed kinetic physics into fluid models, letting them approximate kinetic simulation results without resolving the full phase space. The talk runs 67 minutes in the Bloomberg Lecture Hall and is aimed at listeners already familiar with plasma physics, tracing how different levels of modeling fidelity connect laboratory, space, and astrophysical plasma behavior across scales.