In this study we aim to shed some light into the randomness in the mechanical strength of WC-Co composites at small scale. To do so, computations are performed using the microplane model M7WC for the ceramic phase (WC) and the model MPJ2 for the metallic one (Co). The strength parameter of each individual constitutive phase is determined using a grafted Weibull-Gaussian probability distribution which is sampled using Latin Hypercube algorithm. Models M7WC and MPJ2 have been already calibrated in a recently published study by the authors [1]; thus, these models are assumed to be predictive in this study. The grafted Weibull-Gaussian probability distribution takes into account the randomness due to the (1) crystal orientation of the WC grains, (2) residual stresses that result from material manufacturing processes, and (3) variable level of confinement of Co phases between stiff WC grains. Moreover, as in [1], the meshes are generated using FIB tomography data. In the existing literature, typically a Weibull distribution of strength is assumed at large specimen level. In general, most of these studies have difficulty in identifying the Weibull parameters concisely. In contrast, at small specimen level the existing literature is scarce. Data from tensile tests conducted in WC-Co nanowires as well as from nanoindentation studies of WC-Co composites are used to determine concisely the Weibull parameters along with the grafting point and other statistical parameters of the grafted probability density function. This approach is shown to account successfully, although partly, for the randomness observed in the strength of these ceramic-metal composites.