The latest solar atmosphere models include non-LTE corrections and 3D hydrodynamic convection simulations. These models predict a significant reduction in the solar metal abundance, which leads to a serious conflict between helioseismic data and the predictions of solar interiors models. We demonstrate that the helioseismic constraints on the surface convection zone depth and helium abundance combined with stellar interiors models can be used to define the goodness of fit for a given chemical composition. After a detailed examination of the errors in the theoretical models we conclude that models constructed with the older solar abundances are consistent (<2 sigma) with the seismic data. Models constructed with the proposed new low abundance scale are strongly disfavored, disagreeing at the 15 sigma level. We then use the sensitivity of the seismic properties to abundance changes to invert the problem and infer a seismic solar heavy element abundance mix with two components: meteoritic abundances, and the light metals CNONe. Seismic degeneracies between the best solutions for the elements arise for changes in the relative CNONe abundances and their effects are quantified. We obtain Fe/H=7.50+/-0.045+/-0.003(CNNe) and O/H=8.86+/-0.041+/-0.025(CNNe) for the relative CNNe in the GS98 mixture. The inferred solar oxygen abundance disagree with the abundance inferred from the 3D hydro models. Changes in the Ne abundance can mimic changes in O for the purposes of scalar constraints.Models constructed with low oxygen and high neon are inconsistent with the solar sound speed profile. The implications for the solar abundance scale are discussed.