The predictions of Ahmad-Rolfes-Stepto (ARS) theory, rate theory and Monte-Carlo (MC) modelling of gel points and network moduli at complete reaction are tested against published data from polyurethane-forming polymerizations. Gel points are predicted most accurately by ARS theory and the MC software of Lee and Eichinger. Rate theory and Biosym MC software underestimate extents of intramolecular reaction at gel. Network defects (M(c)/M(c)0) are underestimated by the Lee and Eichinger and Biosym MC software and no distinction between trifunctional (f = 3) and tetrafunctional (f = 4) networks is apparent. Rate theory also underestimates M(c)/M(c)0 but shows a distinction between f = 3 and f = 4 networks. At present, the best approach for predicting M(c)/M(c)0 is : (1) to use ARS theory (or Lee and Eichinger MC software) to predict the gel point; (2) find the effective rate theory ring-forming parameter (lambda(a)0) for that gel point; and (3) use rate theory and that value of lambda(a)0 to determine M(c)/M(c)0.