The role of ruminant liver in modifying the profile of absorbed nutrients before presented to peripheral tissues is illustrated with literature data on hepatic nutrient exchanges. These data are mainly from experiments with multicatheterized dairy cows enabling estimation of blood flow rates and nutrient balances across the liver. In cow liver there is net uptake of propionate, butyrate, valerate, long-chain fatty acids, lactate, amino acids, NH 4 + and O 2, while there is net release of acetate, ketone bodies, glucose, urea and CO 2. Approximately 55% of urea synthesized in the liver is recycled to portal drained viscera. More than 40% of amino acid net absorption is removed by the liver resulting in an amino acid deficit for milk protein synthesis. This could indicate an underestimation of amino acid absorption. Estimated negative carbon balances across the liver can be related to incomplete data concerning lipid exchange and also to carbon use in gluconeogenesis as hepatic uptake of gluconeogenic carbon is less than hepatic release of glucose-carbon. Within a range of glucose flux rates from 3 to 15 mol/d milk yield is linearly related to glucose flux rate. This relationship and published relationships between feed intake and hepatic glucose output are used to calculate regression estimates of basal glucose requirement in cows (2 mol/d) as well as glucose requirement for milk synthesis (0.4 mol/kg milk). Glucose synthesis in the liver is under both metabolic (stimulation by substrate availability; inhibition by glucose itself) and hormonal control (stimulation by glucagon and probably by somatotropin; inhibition by insulin). In early lactation with increased lipid mobilization and fatty acid oxidation in the liver a high rate of gluconeogenesis is ensured by regulation of metabolite transport across mitochondrial membranes. However, in severely underfed animals with excessive lipid mobilization and ketogenesis glucose synthesis is depressed. Reference is given to published mathematical models of liver metabolism in lactating cows. Simulation results obtained with such models have brought further attention to the need for more quantitative knowledge, both regarding hepatic nutrient exchanges (in particular lipids and proteins) and also inter-organ nutrient metabolism (eg. regulation of amino acid catabolism). It is concluded that future feed evaluation systems for ruminants should be based on dynamic, mechanistic whole animal models simulating feed intake, digestion and absorption as well as nutrient metabolism and utilization.