In order to increase milking efficiency, the effects of two different cluster take-off levels (200 and 800 g/min) and feeding vs. not feeding during milking were tested in a Latin square design study including 32 cows. Milk yield, milking time, milk flow and milking interval were measured and milk samples were analysed for gross composition, sodium and potassium concentration, free fatty acid (FFA) content, milk fat globule (MFG) size, MFG membrane (MFGM) material and fatty acid composition. Residual milk was harvested to evaluate udder emptying. Increasing the take-off level from 200 to 800 g/min at the whole udder level decreased milking time and increased harvest flow. Udder emptying decreased slightly, but there were no effects on milk yield, FFA content or MFGM. There were interactive effects of take-off level and feeding during milking on content of fatty acids C4:0, C6:0, C16:0, C18:3(n-3) and C20:0. Feeding during milking increased milk yield per day and decreased milking interval. Sodium and potassium concentrations in milk were unaffected by treatments, indicating no loss of tight junction integrity. From these results, it is clear that feeding during milking should be used to increase milk yield and improve milking efficiency, regardless of take-off level used, and that the effect of feeding is more pronounced when a low take-off level is used. Feeding seemed to counteract the effects of the low take-off level on milking time and milking interval. Low take-off levels can therefore be used in combination with feeding.