Leatherback sea turtles (Dermochelys coriacea) nesting in French Guiana have been the focus of a long-term monitoring and metal flipper-tagging program and were subjected to a double-tagging experiment starting in 1994. We developed a new method to estimate the rate of tag loss based on the number of days at liberty after tagging, using maximum likelihood function fitting. Metal tag loss in leatherback sea turtles is characterized by a high rate of loss just after tagging (>30% of turtles lose at least 1 tag within a year) and is followed by higher retention over a longer time period. Using additional data from permanent internal tags that were subsequently applied to turtles in French Guiana, we have shown that the rate of loss of the second metal tag was higher than that of the first metal tag. More accurate estimates of tag loss rates are essential for correctly generating demographic parameters from Capture–Mark–Recapture (CMR) analyses. We have developed a software package to facilitate the use of these models by others researchers in estimating tag loss.