The association between socioeconomic status (SES) and alcohol-related diseases has been widely explored. Less is known, however, on whether the association of moderate drinking with all-cause mortality is modified by educational level (EL). Using harmonized data from 16 cohorts in the MORGAM Project (N = 142,066) the association of pattern of alcohol intake with hazard of all-cause mortality across EL (lower = primary-school; middle = secondary-school; higher = university/college degree) was assessed using multivariable Cox-regression and spline curves. A total of 16,695 deaths occurred in 11.8 years (median). In comparison with life-long abstainers, participants drinking 0.1–10 g/d of ethanol had 13% (HR = 0.87; 95%CI: 0.74–1.02), 11% (HR = 0.89; 0.84–0.95) and 5% (HR = 0.95; 0.89–1.02) lower rate of death in higher, middle and lower EL, respectively. Conversely, drinkers > 20 g/d had 1% (HR = 1.01; 0.82–1.25), 10% (HR = 1.10; 1.02–1.19) and 17% (HR = 1.17; 1.09–1.26) higher rate of death. The association of alcohol consumption with all-cause mortality was nonlinear, with a different J-shape by EL levels. It was consistent across both sexes and in various approaches of measuring alcohol consumption, including combining quantity and frequency and it was more evident when the beverage of preference was wine. We observed that drinking in moderation (≤ 10 g/d) is associated with lower mortality rate more evidently in individuals with higher EL than in people with lower EL, while heavy drinking is associated with higher mortality rate more evidently in individuals with lower EL than in people with higher EL, suggesting that advice on reducing alcohol intake should especially target individuals of low EL.