Nanoparticles dispersed in polymer melts have recently been shown to decrease the bulk viscosity. This contradicts expectations based on Einstein's well-known theory for effective viscosity of dilute, random dispersions of rigid spheres in Newtonian fluids. In this paper, we examine a continuum hydrodynamic model where a layer of polymer at the nanoparticle–polymer interface has a different viscosity and density than the bulk polymer. When the layer thickness is greater than the nanoparticle radius, and the layer viscosity is lower than that of the bulk polymer, the intrinsic viscosity is comparable to the unexpectedly large, negative values reported experimentally by Mackay and coworkers. Accordingly, our continuum hydrodynamic model attributes a bulk viscosity reduction to a lower melt viscosity at the nanoparticle–polymer interface. Such a reduction has been ascribed to the Rouse dynamics of polymer chains at the nanoparticle–polymer interface. Fitting the theory to experimental data reveals a simple correlation between the reduced viscosity layer thickness, polymer chain size, entanglement tube diameter, and nanoparticle size and concentration. Our calculations support the arguments of Mackay and coworkers that nanoparticle inclusions significantly influence the polymer chain conformation in melts, even when the inclusion volume fraction is very small.