Through coarse-grained molecular dynamics simulations, the effects of nanoparticle properties, polymer–nanoparticle interactions, chain crosslinks and temperature on the stress–strain behavior and mechanical reinforcement of polymer nanocomposites (PNCs) are comprehensively investigated. By regulating the filler–polymer interaction (miscibility) in a wide range, an optimal dispersion state of nanoparticles is found at moderate interaction strength, while the mechanical properties of PNCs are improved monotonically with the increase of the particle–polymer interaction due to the tele-bridge structures of nanoparticles via polymer chains. Although smaller-sized fillers more easily build interconnected structures, the elastic moduli of PNCs at the percolation threshold concentration where a three-dimensional filler network forms are almost independent of nanoparticle size. Compared with spherical nanoparticles, anisotropic rod-like ones, especially with larger aspect ratio and rod stiffness, contribute exceptional reinforcement towards polymer materials. In addition, the elastic modulus with the strain, derived from the stress–strain curve, shows an analogous nonlinear behavior to the amplitude-dependence of the storage modulus (Payne effect). Such a behavior originates essentially from the failure/breakup of the microstructures contributing to the mechanical reinforcement, such as bound polymer layers around nanoparticles or nanoparticle networking structures. The Young's modulus as a function of the nanoparticle volume fraction greatly exceeds that predicted by the Einstein–Smallwood model and Guth–Gold model, which arises primarily from the contribution of the local/global filler network. The temperature dependence of the Young's modulus is further examined by mode coupling theory (MCT) and the Vogel–Fulcher–Tammann (VFT) equation, and the results indicate that the time–temperature superposition principle holds modestly above the critical temperature on the short-time (small-length) scale of elastic response. This work is expected to provide some guidance on controlling and improving the mechanical properties and nonlinear behavior of PNCs.