Five molecular models of rubber elasticity which employ different treatments of entanglement effects (the Kloczkowski-Mark-Erman diffused-constraint model, the Edwards-Vilgis (E-V) slip-link model, the tube models of Gaylord-Douglas (G-D), Kaliske-Heinrich, Rubinstein-Panyukov versions) are assessed using biaxial deformation data for an entanglement-dominated network of end-linked poly(dimethylsiloxane) (PDMS) in which trapped entanglements are dominant in number relative to chemical cross-links. The theoretical stress-strain relations were calculated from the elastic free energy (M of each model. Using the reduced stress (the nominal stress divided by equilibrium modulus G(o)), the strain-dependent predictions of each model were tested from two different viewpoints, i.e., the dependence of the reduced stresses on the principal ratio and the I-i dependence of (partial derivativeW/partial derivativeI(j))/G(o) (i,j = 1,2), where I-1 and I-2 are the first and second invariants of deformation tensor (the Rivlin-Saunders method). The diffused-constraint model is relatively successful in reproducing the reduced stress-strain data over a wide range of deformations, but the model definitely underestimates the magnitude of G because it does not consider trapped entanglements as additional cross-links contributing to G., in contrast to the tube models and the slip-link models. The G-D tube model is more successful in reproducing the experimental data relative to the other two versions of the tube model, but the G-D model obviously underestimates the stresses at large deformations. Among the five molecular theories tested here, the E-V slip-link model shows the most successful reproducibility over large portions of the experimental results. The agreements in reduced stress-strain relations are satisfactory over the entire deformation range, although considerable disagreement is recognized in the I-i dependence of partial derivativeW/partial derivativeI(2). Also, the fitted parameter values in the E-V slip-link model are fairly well explained using the molecular considerations based on the structural characteristics of the network sample employed here.