continental-scale field experiment conducted over 3 years, in which the amount of total nitrogen yield (Ntot) and the gain of N yield in mixtures as compared to grass monocultures (Ngainmix) was quantified from four-species grass–legume stands with greatly varying legume proportions. Stands consisted of monocultures and mixtures of two N2-fixing legumes and two nonfixing grasses. The amount of Ntot of mixtures was significantly greater (P ≤ 0.05) than that of grass monocultures at the majority of evaluated sites in all 3 years. Ntot and thus Ngainmix increased with increasing legume proportion up to one-third of legumes. With higher legume percentages, Ntot and Ngainmix did not continue to increase. Thus, across sites and years, mixtures with one-third proportion of legumes attained ~95% of the maximum Ntot acquired by any stand and had 57% higher Ntot than grass monocultures. Realized legume proportion in stands and the relative N gain in mixture (Ngainmix/Ntot in mixture) were most severely impaired by minimum site temperature (R = 0.70, P = 0.003 for legume proportion; R = 0.64, P = 0.010 for Ngainmix/Ntot in mixture). Nevertheless, the relative N gain in mixture was not correlated to site productivity (P = 0.500), suggesting that, within climatic restrictions, balanced grass–legume mixtures can benefit from comparable relative gains in N yield across largely differing productivity levels. We conclude that the use of grass–legume mixtures can substantially contribute to resource-efficient agricultural grassland systems over a wide range of productivity levels, implying important savings in N fertilizers and thus greenhouse gas emissions and a considerable potential for climate change mitigation.