Until recently, the simulation of transition metal particles in the nanometer range was only feasible with semi-empirical approaches and classical molecular dynamics simulations. However, the close interrelation of electronic and structural properties often leaves no alternative to a fully quantum mechanical treatment. The evolution of modern supercomputer technology nowadays allows the simulation of nanometer-sized objects from first principles in the framework of density functional theory (DFT). A technologically relevant example is the search for ultra-high density magnetic recording media where the decrease of the magnetic grain size competes with the onset of superparamagnetism. Here, Fe-Pt nanoparticles are discussed as a promising solution to the problem due to their large magnetocrystalline anisotropy in the ordered L10 phase. However, in experiment also other, less favorable, structures are observed. Therefore, a systematic ab initio investigation of the morphologies of transition metal nanoparticles with respect to their energetics and magnetism appears highly desirable. Within this contribution, we discuss the results of recent DFT calculations of Fe and Fe-Pt clusters with up to 561 atoms including full geometric optimization.