Two experiments investigated two-year-olds' retention and generalization of novel words across short and long time delays. Specifically, retention of newly learned words and generalization to novel exemplars or novel contexts were tested 1 min or 1 week after learning. Experiment 1 revealed successful retention as well as successful generalization to both new exemplars and new contexts after a one-minute delay, with no statistical differences between retention and generalization performance for either generalization type. Toddlers tested after a week delay (Experiment 2) showed successful retention and generalization as well, but while context generalization was statistically equivalent to retention accuracy, exemplar generalization was significantly lower than retention accuracy. The overall success in both retention and generalization suggests that toddlers' newly learned words are robust and flexible. However, the lower exemplar generalization performance compared to retention after a weeklong delay suggests that novel words may become less flexible across exemplar characteristics over time.