Many biological and engineering materials are made from fibers organized in the form of a stochastic crosslinked network, and the mechanics of the network controls the behavior of the material. In this work we investigate the strength of stochastic networks without pre-existing damage which fail due to crosslink rupture. Athermal networks ranging from approximately affine to strongly non-affine are subjected to multiaxial loading and the strength is evaluated using numerical models. It is observed that once the stress is normalized by the strength measured in uniaxial tension, the failure surface becomes approximately independent of network parameters. This extends the relation between strength and network parameters previously established in (S. Deogekar, M. R. Islam, R. C. Picu, Parameters controlling the strength of stochastic fibrous materials, Int. J. Solids Struct., 2019, 168, 194–202) to the multiaxial case. The failure surface depends on both first two invariants of the stress. Strongly non-affine networks behave somewhat different from the affine networks under loadings close to the hydrostatic and pure shear loading modes, while the difference disappears in the first quadrant of the principal stress space. The results are compared with experimental data from the literature.