The aim of our study was to investigate the effect of caffeine supplementation on 3-min all-out arm crank exercise performance in paraplegic (P) and tetraplegic (T) compared with able-bodied (AB) participants. A placebo-controlled, randomized, crossover, and double-blind study design was chosen to investigate the differences between caffeine (CAF) and placebo (PLC). In total, 34 healthy, trained participants were tested. Seventeen were AB (median [minimum; maximum] VO2peak: 33.9 mL/min/kg [23.6; 57.6]), 10 were P (VO2peak: 34.4 mL/min/kg [19.5; 48.8]), and 7 were T (VO2peak: 13.6 mL/min/kg [8.6; 16.3]). All participants performed two 3-min all-out tests on an arm crank ergometer following the ingestion of either PLC or CAF. Power output parameters, plasma caffeine (PC), epinephrine (EPI), and norepinephrine (NOR) concentrations were assessed. CAF significantly increased average power over the first 30 s (p = .028) and 60 s (p = .005) in P, but not in T (p = .61; p = .87) nor in AB (p = .25; p = .44). Peak power was increased in the CAF trial in AB (+46 W) as well as in P (+21 W) but was not significantly different from PLC (AB: p = .10; P: p = .17). PC significantly increased in all groups (AB: p = .002; P: p = .005; T: p = .018) whereas EPI showed a significant increase only in AB (p = .002) and in P (p = .018). NOR increased significantly in AB (p = .018) but did not increase in the other groups. Caffeine seems to enhance short-duration exercise performance in P. In contrast, T showed a high interindividual variability and overall no ergogenic effect was detected in this group.