The processes controlling preservation and recycling of particulate biogenic silica in sediments must be understood in order to calculate oceanic silica mass balances. The new contribution of this work is the coupled use of advanced models including reprecipitation and different phases of biogenic silica with different dissolution characteristics as well as new data sets concerning experimental dissolution rates of marine particles and sediment opal obtained in flow-through reactors. We have used three models representing early diagenesis of biogenic silica in sediments that calculate the vertical distributions of dissolved silicate and solid silica in sediments. Model 1 contains one type of biogenic silica and the dissolution rate is constant, whereas model 2 contains a variable dissolution rate constant with sediment depth (representing aging) and one type of biogenic silica. Model 3 incorporates aging by describing two types of biogenic silica that differ by their dissolution properties. An explicit term of reprecipitation is incorporated into the three models. The distributions of dissolved silicate and solid silica predicted by steady-state calculations are compared to 4 observed data sets, from the Southern Ocean, the Equatorial Pacific and the North Atlantic, covering a wide range of sediment compositions, from opal-rich to opal-poor sediments. After adjustment of the critical parameters (the apparent silica dissolution rate constants, the biogenic silica flux deposited at the sediment–water interface and the reprecipitation rate), the second and third models provide good agreements between predicted and measured dissolved Si and solid silica profiles for each data set, except for the second model in the Equatorial Pacific. However, a large discrepancy between the experimentally derived dissolution rate constants and those calculated by the models is observed at all sites at depth in the sediment, suggesting that either artifacts arise during dissolution experiments such as over-representation of rapidly dissolving silica or variation of dissolution properties during the experimental procedures or the model oversimplifies the processes associated to silica dissolution and alumino-silicate reprecipitation and their interaction. ; The processes controlling preservation and recycling of particulate biogenic silica in sediments must be understood in order to calculate oceanic silica mass balances. The new contribution of this work is the coupled use of advanced models including reprecipitation and different phases of biogenic silica with different dissolution characteristics as well as new data sets concerning experimental dissolution rates of marine particles and sediment opal obtained in flow-through reactors. We have used three models representing early diagenesis of biogenic silica in sediments that calculate the vertical distributions of dissolved silicate and solid silica in sediments. Model 1 contains one type of biogenic silica and the dissolution rate is constant, whereas model 2 contains a variable dissolution rate constant with sediment depth (representing aging) and one type of biogenic silica. Model 3 incorporates aging by describing two types of biogenic silica that differ by their dissolution properties. An explicit term of reprecipitation is incorporated into the three models. The distributions of dissolved silicate and solid silica predicted by steady-state calculations are compared to 4 observed data sets, from the Southern Ocean, the Equatorial Pacific and the North Atlantic, covering a wide range of sediment compositions, from opal-rich to opal-poor sediments. After adjustment of the critical parameters (the apparent silica dissolution rate constants, the biogenic silica flux deposited at the sediment–water interface and the reprecipitation rate), the second and third models provide good agreements between predicted and measured dissolved Si and solid silica profiles for each data set, except for the second model in the Equatorial Pacific. However, a large discrepancy between the experimentally derived dissolution rate constants and those calculated by the models is observed at all sites at depth in the sediment, suggesting that either artifacts arise during dissolution experiments such as over-representation of rapidly dissolving silica or variation of dissolution properties during the experimental procedures or the model oversimplifies the processes associated to silica dissolution and alumino-silicate reprecipitation and their interaction.