Electrical machines with stranded random windings often suffer from considerable circulating current losses. These losses have been poorly studied because of the difficulty and computational cost of modeling stranded windings, and the stochastic nature of the problem due to the uncertain positions of the strands. This paper proposes two methods to model random stranded windings of arbitrary complexity. First, a circuit model considering the entire main flux path is presented, and some practical implementation considerations are discussed. Second, a computationally efficient finite-element approach based on non-conforming meshing is presented. Finally, a method is proposed to model the random packing process of strands within a slot, without any remeshing or inductance recalculation required. The proposed methods are then compared with special no-rotor measurement data of a large number of high-speed induction machines, and good agreement is observed.