OBJECTIVE The management of upper gastrointestinal bleeding requires training of the endoscopist. We aimed to validate a live animal model of bleeding ulcers for training in endoscopic hemostasis. METHODS Bleeding ulcers were created by repeated grasp-and-snare gastric mucosectomies in pigs rendered "bleeders" by preadministration of clopidogrel, aspirin, and unfractionated heparin. The feasibility and reproducibility of the model (proportion of bleeding ulcers, number of ulcers per animal, and time needed to produce a bleeding ulcer) were prospectively evaluated in six animals. Ten endoscopic experts assessed the similarity of this pig model to human bleeding ulcers (four-point Likert scale). The training capabilities of the model for hemostatic techniques (needle injection, bipolar electrocoagulation, and hemoclipping) were evaluated in 46 fellows (four-point Likert scale). RESULTS A total of 53 gastric ulcers were created in 6 animals (8.8 ± 1.5 ulcers/animal). Successful active ulcer bleeding (Forrest Ib) was achieved in 96.2 % of cases. Bleeding was moderate to abundant in 79 % of cases. Ulcerations consistently reached the submucosal layer. The mean (± SD) time taken to create a bleeding ulcer was 3.8 ± 0.6 minutes. Endoscopic experts assessed the realism of the ulcers and bleeding at 3.2 ± 0.7 and 3.6 ± 0.7 respectively on a four-point Likert scale. The training significantly improved the endoscopic skills of the 46 fellows (P < 0.0001) in all hemostatic techniques. CONCLUSIONS The live porcine model of bleeding ulcers was demonstrated to be realistic, reproducible, feasible, time efficient, and easy to perform. It was favorably assessed as an excellent model for training in endoscopic treatment of bleeding ulcers.