Abstract Background Live tissue models are considered the most useful simulation for training in the management for hemostasis of penetrating injuries. However, these models are expensive, with limited opportunities for repetitive training. Ex-vivo models using tissue and a fluid pump are less expensive, allow repetitive training and respect ethical principles in animal research. The purpose of this study is to objectively evaluate the effectiveness of ex-vivo training with a pump, compared to live animal model training. Staff surgeons and residents were divided into live tissue training and ex-vivo training groups. Training in the management of a penetrating cardiac injury was conducted for each group, separately. One week later, all participants were formally evaluated in the management of a penetrating cardiac injury in a live animal. Results There are no differences between the two groups regarding average years of experience or previous trauma surgery experience. All participants achieved hemostasis, with no difference between the two groups in the Global Rating Scale score (ex-vivo: 25.2 ± 6.3, live: 24.7 ± 6.3, p = 0.646), blood loss (1.6 ± 0.7, 2.0 ± 0.6, p = 0.051), checklist score (3.7 ± 0.6, 3.6 ± 0.9, p = 0.189), or time required for repair (101 s ± 31, 107 s ± 15, p = 0.163), except overall evaluation (3.8 ± 0.9, 3.4 ± 0.9, p = 0.037). The internal consistency reliability and inter-rater reliability in the Global Rating Scale were excellent (0.966 and 0.953 / 0.719 and 0.784, respectively), and for the checklist were moderate (0.570 and 0.636 / 0.651 and 0.607, respectively). The validity is rated good for both the Global Rating Scale (Residents: 21.7 ± 5.6, Staff: 28.9 ± 4.7, p = 0.000) and checklist (Residents: 3.4 ± 0.9, Staff Surgeons: 3.9 ± 0.3, p = 0.003). The results of self-assessment questionnaires were similarly high (4.2–4.9) with scores in self-efficacy increased after training (pre: 1.7 ± 0.8, post: 3.2 ± 1.0, p = 0.000 in ex-vivo, pre: 1.9 ± 1.0, post: 3.7 ± 0.7, p = 0.000 in live). Scores comparing pre-training and post-evaluation (pre: 1.7 ± 0.8, post: 3.7 ± 0.9, p = 0.000 in ex-vivo, pre: 1.9 ± 1.0, post: 3.8 ± 0.7, p = 0.000 in live) were increased. Conclusion Training with an ex-vivo model and live tissue training are similar for the management of a penetrating cardiac injury, with increased self-efficacy of participants in both groups. The ex-vivo model is useful to learn hemostatic skills in trauma surgery.