Braided composites combine high structural stability with low cost, exceptional damage tolerance and attractive energy absorption due to yarn interlacing. The rapid growth of braided applications requires well developed failure predictions, which can accurately calculate material behaviors and specifically the initiation and propagation of damage. However, failure and strength analysis is more demanding than stiffness prediction, since nonlinearities in material models and contact assumptions have to be introduced, along with failure criteria and material degradation models. Uncertainty assessment approaches have not been applied to strength predictions yet, as there are no studies established on a probabilistic framework. This study attempts a multiscale modeling framework, accounting for a variety of uncertainties, under static tensile loading states. A numerical microscale model was established for the strength properties prediction of the yarns, under different loading conditions. 3D Hashin and modified von Mises (Christensen) failure criteria were applied to the mesoscale model, along with a stiffness degradation scheme in a user-defined subroutine for ABAQUS. A variety of random parameters are introduced to the system, covering a wide range of material and geometrical uncertainties, caused by either lack of knowledge or manufacturing processes. The response variability of the macroscale strength properties is calculated within a Monte Carlo framework. Some numerical aspects are also discussed, since all algorithms must be kept robust and computationally effortless, due to the probabilistic nature of the models. When it comes to the variability influence and sensitivity of random parameters, efficient surrogate modeling techniques are proposed, in order to overcome the computational burden of the FE discretization and the curse of dimensionality. The study is expected to provide a first insight towards stochastic analysis in terms of strength prediction for braided composites, but also highlights the importance of realistic uncertainty quantification.