Model projections of ice flow in a changing climate are dependent on model inputs such as surface elevation, bedrock position or surface temperatures, among others. Of all these inputs, geothermal heat flux is the one for which uncertainty is greatest. In the area of Pine Island Glacier, Antarctica, available data sets differ by up to a factor of 2.5. Here, we evaluate the impact of such uncertainty on ice flow, using sampling analyses based on the Latin-Hypercube method. First, we quantify the impact of geothermal heat flux errors on ice hardness, a thermal parameter that critically controls the magnitude of ice flow. Second, we quantify the impact of the same errors on mass balance, specifically on the mass flux advecting through thirteen fluxgates distributed across Pine Island Glacier. We contrast our results with similar uncertainties generated by errors in the specification of ice thickness. Model outputs indicate that geothermal heat flux errors yield uncertainties on ice hardness on the order of 5–7%, with maximum uncertainty reaching 15%. Resulting uncertainties in mass balance remain however below 1%. We discuss the uncertainty distribution and its relationship to the amount of heat available at the base of the ice sheet from friction, viscous and geothermal heating. We also show that comparatively, errors in ice thickness contribute more to model uncertainty than errors in geothermal heat flux, especially for fast-flowing ice streams.