Because athletic horses become hypoxaemic and hypercapnoeic during high-intensity exercise but ponies do not, six Thoroughbred horses and five ponies performed an incremental exercise test at speeds with calculated energy requirements that were 40, 60, 80 and 115% of V˙O2max, with the objective of comparing their blood gas and ventilatory responses to exercise. Expired gas and blood samples were taken and breathing mechanics were assessed before exercise and during the last 15 s at each intensity. Maximal V˙O2 and V˙CO2 in horses were 153±5 (SEM) and 187±4 ml kg−1 min−1, respectively, while corresponding values in ponies were 92±4 and 112±7 ml kg−1 min−1. During heavy and supramaximal exercise, horses, but not ponies, became hypoxaemic and hypercapnic. There was no significant difference for V˙E kg−1 between groups during maximal exercise, but PAO2, PaO2 and PvO2 were lower and PaCO2 and [(A−a)O2D] were greater in horses than in ponies. Additionally, the horses' maximal transpulmonary pressure difference was higher and their total pulmonary resistance and ventilatory equivalent lower than in ponies. Flow-volume loops suggested that horses experienced expiratory flow limitation but that ponies did not. These results indicated that horses like Thoroughbreds appear to be expiratory flow-limited and become hypoxaemic and hypercapnic when the demand for gas exchange associated with their high V˙O2max and V˙CO2max is greater than can be met by their ventilatory system. Ponies, which are less capable athletes, could better match their ventilatory response with their metabolic capabilities and so were able to maintain PaO2 in the pre-exercise range and decrease PaCO2 to a tension that was more compatible with acid–base homeostasis.