OBJECTIVE For optimizing the quality of repeated-sprint training in hypoxia, the differences in the acute performance responses to a single session of repeated-sprint exercise with various (i) inspired oxygen fractions; (ii) exercise-to-recovery (E:R) ratios and (iii) recovery modalities were examined. METHODS Ten male participants performed three sets, 5 × 5-s all-out treadmill sprints, E:R ratio of 1:5, passive recovery, in seven trials randomly. In four of the seven trials, hypoxic levels were set corresponding to sea level (SL1:5P), 1500 (1.5K1:5P), 2500 (2.5K1:5P), and 3500 m (3.5K1:5P), respectively. In a further two trials, the hypoxic level of 3.5K1:5P was maintained, while the E:R ratio was reduced to 1:4 (3.5K1:4P) and 1:3 (3.5K1:3P), respectively. In the last trial, the passive recovery mode of 3.5K1:5P was changed to active (3.5K1:5A). RESULTS In comparison to SL1:5P, the averaged peak velocity (P-Vel), mean velocity (M-Vel), and velocity decrement score (Sdec) of the sprints, and the cumulative HR-based training impulse (cTRIMP) in 1.5K1:5P and 2.5K1:5P were well maintained. Minor decrement in the M-Vel was found in 3.5K1:5P. Conversely, lowered E:R ratio in 3.5K1:4P and 3.5K1:3P significantly reduced the P-Vel (≥ -2.3%, Cohen's d ≥ 0.43) and M-Vel (≥ -2.4%, ≥ 0.49), and in 3.5K1:3P altered the Sdec (107%, ≥ 0.96), and cTRIMP (-16%, 1.39), when compared to 3.5K1:5P. Furthermore, mild reductions in M-Vel (-2.6%, 0.5) was observed in 3.5K1:5A using the active recovery mode. Other variables did not change. CONCLUSIONS The findings suggest that a 3.5K1:5P marginally maintained sea-level training loads, and as a result, could maximally optimize the training stress of hypoxia.