This study tested the effects of low-cadence (60 rev min−1) uphill (Int60) or high-cadence (100 rev min−1) level-ground (Int100) interval training on power output (PO) during 20-min uphill (TTup) and flat (TTflat) time-trials. Eighteen male cyclists ( $$ \dot{V}{\text{O}}_{2\max } $$ : 58.6 ± 5.4 mL min−1 kg−1) were randomly assigned to Int60, Int100 or a control group (Con). The interval training comprised two training sessions per week over 4 weeks, which consisted of six bouts of 5 min at the PO corresponding to the respiratory compensation point (RCP). For the control group, no interval training was conducted. A two-factor ANOVA revealed significant increases on performance measures obtained from a laboratory-graded exercise test (GXT) (P max: 2.8 ± 3.0%; p < 0.01; PO and $$ \dot{V}{\text{O}}_{2} $$ at RCP: 3.6 ± 6.3% and 4.7 ± 8.2%, respectively; p < 0.05; and $$ \dot{V}{\text{O}}_{2} $$ at ventilatory threshold: 4.9 ± 5.6%; p < 0.01), with no significant group effects. Significant interactions between group and uphill and flat time-trial, pre- versus post-training on PO were observed (p < 0.05). Int60 increased PO during both TTup (4.4 ± 5.3%) and TTflat (1.5 ± 4.5%). The changes were −1.3 ± 3.6, 2.6 ± 6.0% for Int100 and 4.0 ± 4.6%, −3.5 ± 5.4% for Con during TTup and TTflat, respectively. PO was significantly higher during TTup than TTflat (4.4 ± 6.0; 6.3 ± 5.6%; pre and post-training, respectively; p < 0.001). These findings suggest that higher forces during the low-cadence intervals are potentially beneficial to improve performance. In contrast to the GXT, the time-trials are ecologically valid to detect specific performance adaptations.