Purpose To determine embryo and cycle-specific parameters associated with twin live birth in day 3 cryopreserved double embryo transfer (DET) cycles, and to propose a new prediction model for external validation. Methods All cycles with autologous or donor cryopreserved DET of day 3 embryos from 2002 to 2011 at a single academic institution with a singleton or twin live birth were included (n=207). Patient characteristics, post-thaw embryo morphology and freeze-thaw parameters were compared between patients with a single and twin live birth. Results Demographic characteristics were similar, except that patients delivering twins were younger at age of cryopreserved embryo transfer (CET), fewer were parous and more were anovulatory. Duration of embryo storage, time in culture post-thaw, endometrial thickness and use of assisted hatching were comparable. Six predictors of twin delivery were identified: patient age <35 year at CET, intact survival of the lead embryo, resumption of mitosis, 7-8 viable cells in the non-lead embryo, transfer of a lead embryo with ≥7 cells and a sum of ≥14 viable cells in the two transferred embryos. Regression modeling predicted a step-wise increase in the probability of twins with addition of each predictor; with all six present, the risk of twins was predicted to be 53 % and with none present, the risk decreased to 6 %. Conclusions The six identified variables associated with twin live birth following day 3 cryopreserved DET have been applied to derive a prediction model for estimating the risk of twin delivery. External validation of the model is required prior to clinical application.