Linear elastic fracture modeling coupled with empirical material tension data result in good quantitative agreement with experimental measurements of fracture failure for both brittle and tough epoxy nanocomposites. The nanocomposites comprise diglycidyl ethers of bisphenol A cured with O,O’ bis (2-aminopropylpropylene glycol) (Jeffamine D230) and doped with rubber nanoparticles of varying concentrations. Toughness, critical load, and critical displacement in quasi-static single edge-notched three-point bending are predicted accurately using both surface-based cohesive zone (CZS) and extended finite element (XFEM) methods implemented in Abaqus software. Fracture initiation within a crack is taken at the yield stress from uniaxial tension data. Prediction of fracture processes using a generalized truncated linear traction-separation law was significantly improved by considering the case of a linear softening function. There are no adjustable parameters in the XFEM model. The CZS model requires only optimization of the element displacement at fracture parameter. Thus, these continuum methods describe these materials in mode I fracture with a minimum number of independent parameters.