A germplasm collection of 484 accessions of Vicia faba was screened for resistance to rust (Uromyces viciae-fabae) under field conditions. Accessions varied in the levels of rust infection, although no complete resistance was identified. Stability of resistance of the 39 most-resistant accessions was tested in a multi-location experiment in Austria, Egypt, Tunisia, United Kingdom and Spain over three additional field seasons. Genotype × environment interaction accounted for 43% of the sum of squares of the multi-environment evaluation, revealing instability of the phenotypic expression across environments. This might hamper the efficiency of selection suggesting the need for selection in different environments. Three possible mega-environments were discerned in the studied area, Mediterranean (Spain, Tunisia and Egypt), Oceanic (UK) and Continental (Austria). Córdoba (Spain) and Kafr El-Sheik (Egypt) showed as ideal environments for rust resistance screenings within Mediterranean environment. Several accessions (300, 303, 311, 313, 720, 1196 and 1271) were grouped as moderately to highly resistant in the three defined mega-environments. These accessions showed clear differences both in terms of reduced disease severity and high stability, which make them good candidates for international faba bean breeding programmes. Concerning each mega-environment, accessions 300 and 311 were the most resistant and stable ones across the Mediterranean one, followed by accessions 720, 1022, 1272, 1320 and BPL261. On the contrary other accessions (313, 452, 481 and 1196) were the most resistant in Oceanic and Continental environments. However, 452 and 481 were susceptible in the Mediterranean mega-environment. This contrasting performance across the environments was also supported by contradictory performance of the checks BPL261 and Baraca in Oceanic and Continental environments, suggesting differential virulence in rust populations, which deserves further attention.