The heat and salt fluxes produced by salt fingering at a density interface are studied with a numerical and an analytical model. Specifically, the issue concerning the value of the heat‐to‐salt flux ratio is addressed. The numerical modeling based on direct numerical computation of the nonlinear governing equations obtains values around 0.5. This value is approximately the average of widely varying experimental values reported in the literature. The large difference between the theoretical flux ratio predicted based on the buoyancy maximization hypothesis and the experimentally derived flux ratio is examined with an analytical model, which includes both effects of salt stratification in the interface and salt discontinuity at the edges of the interface. Combined with the numerical model results, the analysis shows that the disagreement can be traced to the flux maximization hypothesis itself. An alternative hypothesis that maximizes convective velocity amplitude is presented which gives flux‐ratio predictions consistent with the measurements. Discussion of the finger width scale and the magnitude of the salt flux is also given as well as comparison of the salt flux with laboratory measurements.