The deformation mechanics of dry networks of large-aspect-ratio fibers with random orientation controls the processing of long-fiber thermoplastics (LFTs) and greatly affects the mechanical properties of the final composites. Here, we generate initial geometries of fiber networks in a cubic unit cell with a fiber aspect ratio of l/d = 100 and fully periodic boundary conditions for later numerical simulation. The irreversible random sequential adsorption (RSA) process is first used to generate a quasi-random structure due to the excluded-volume requirements. In order to investigate the nonequilibrium character of the RSA, a second method, which is similar to the mechanical contraction method (MCM) (Williams and Philipse, 2003, “Random Packings of Spheres and Spherocylinders Simulated by Mechanical Contraction,” Phys. Rev. E, 67, pp. 1–9) and based on a simplified Metropolis Monte Carlo (MC) simulation is then developed to produce quasi-equilibrium fiber geometries. The RSA packing results (ϕ ≈ 4.423% when using a fiber aspect ratio of 100) are in good agreement with the maximum unforced random packing limits (Evans and Gibson, 1986, “Prediction of the Maximum Packing Fraction Achievable in Randomly Oriented Short-Fibre Composites,” Compos. Sci. Technol., 25, pp. 149–162). The fiber structures were characterized by several distribution functions, including pair-spatial and pair-orientation distributions, based on either the center-to-center distance or the shortest distance between the particles. The results show that the structures generated by the RSA have an easily-detectable long-range spatial correlation but very little orientational correlation. In contrast, the quasi-equilibrium structures have reduced spatial correlation but increased short-range orientational correlation.