The purpose of this study was to compare cardiopulmonary, neuromuscular, and metabolic responses to severe-intensity intermittent exercises with variable or constant work rate. Eleven cyclists (28 ± 5 years; 74 ± 7 kg; 175 ± 5 cm; 63 ± 4 mL·kg-1·min-1) performed the following tests until exhaustion on separate days: 1) an incremental test; 2) in random order, two constant work rate tests at 95% and 110% of the peak power for the determination of Critical Power (CP); 3) 2–4 tests for the determination of the highest power that still permits the achievement of maximal oxygen uptake (PHIGH); and 4) two random severe-intensity intermittent exercises. The last two sessions consisted of a constant work rate (CWR) exercise performed at PHIGH or a decreasing work rate (DWR) exercise from PHIGH until 105% of CP. Compared to CWR, DWR presented higher time to exhaustion (635 ± 223 vs. 274 ± 65 s), time spent above 95% of VO2max (t95%VO2max) (323 ± 227 vs. 98 ± 65 s), and O2 consumed (0.97 ± 0.41 vs. 0.41 ± 0.11 L). Electromyography amplitude (RMS) decreased for DWR but increased for CWR during each repetition. However, RMS and VO2 divided by power output (RMS/PO and VO2/PO ratio) increased in every repetition for both protocols, but to a higher extent and slope for DWR. These findings suggest that the higher RMS/PO and VO2/PO ratio in association with the longer exercise duration seemed to have been responsible for the higher t95%VO2max observed during severe DWR exercise.