Simulations of sheet metal forming processes are nowadays of common use in the manufacturing industry and mostly performed with commercial finite-element softwares. Though remarkable improvements on the reliability of these simulations have been achieved over the last decades, the material modeling remains an essential aspect to be upgraded. On one hand, the constitutive laws available in these commercial softwares are phenomenological by nature as they are based on discrete experimental observations. However, industrial forming simulations involve complex straining paths, often leading to material behaviors that are out of the range of the experimental data used for building and fitting these phenomenological models thus limiting the predictiveness of the simulations. These effects are of critical importance especially in the field of micro-forming where parts are obtained from ultra-thin sheets metals whose behavior can be highly heterogeneous as a result of a small number of grains in their thickness, local deformations becoming dominant. On the other hand, Crystal Plasticity Finite Element Models (CPFEM) are based on the physical crystallographic slip that occurs on crystalline material slip planes during straining. This kind of models relate the material behavior to its microstructure, therefore allowing for a deep representation of the physics of deformation. This work aims to establish a one-to-one comparative study of these two modeling approaches. It is highlighted that, contrariwise to phenomenological models, CPFEM are less subordinated to the broadness and the quality of the experimental database used for their adjustment and are expected to describe accurately the material behavior under several strain paths. It is also shown that, provided an efficient and robust implementation taking advantage of parallel computing, CPFEM represent a workable modeling basis for simulations. Then, applications to the simulation of two industrial micro-forming processes with ultra-thin copper based alloys of different thickness to grain size ratios are conducted. The objective is to provide a guidance tool to assess in which cases phenomenological models based simulations are still relevant or when one should rather switch to CPFEM.