The influence of training status on the maximal accumulated oxygen deficit (MAOD) was used to assess the validity of the MAOD method during supra-maximal all-out cycle exercise. Sprint trained (ST; n = 6), endurance trained (ET; n = 8), and active untrained controls (UT; n = 8) completed a 90 s all-out variable resistance test on a modified Monark cycle ergometer. Pretests included the determination of peak oxygen uptake (VO2peak) and a series (5-8) of 5-min discontinuous rides at submaximal exercise intensities. The regression of steady-state oxygen uptake on power output to establish individual efficiency relationships was extrapolated to determine the theoretical oxygen cost of the supramaximal power output achieved in the 90 s all-out test. Total work output in 90 s was significantly greater in the trained groups (P < 0.05), although no differences existed between ET and ST. Anaerobic capacity, as assessed by MAOD, was larger in ST compared to ET and UT. While the relative contributions of the aerobic and anaerobic energy systems were not significantly different among the groups, ET were able to achieve significantly more aerobic work than the other two groups, while ST were able to achieve significantly more anaerobic work. Peak power and peak pedalling rate were significantly higher in ST. The results suggested that MAOD determined during all-out exercise was sensitive to training status and provided a useful assessment of anaerobic capacity. In our study sprint training, compared with endurance training, appeared to enhance significantly power output and high intensity performance over brief periods (up to 60 s), yet few overall differences in performance (i.e. total work) existed during 90 s of all-out exercise.