The purpose of this study was to confirm reduced training metrics previously associated with a ketogenic low-carbohydrate (CHO) high-fat diet (LCHF) and investigate their attenuation with caffeine supplementation. At baseline, n=21 elite race-walkers followed a high CHO availability (HCHO) diet and performed a tempo hill session (14 km with a 450 m elevation gain). Athletes were then assigned to either the HCHO or LCHF in a parallel groups design for 3 weeks where the 14 km tempo hill session was repeated each week. On weeks 2 and 3, in a randomized crossover allocation, all participants received 3 mg/kg caffeine or placebo (gum), 20 min before the session. Race-walking speed, heart rate, RPE, blood metabolites and Stroop word-color test metrics were collected. Although LCHF athletes walked faster at baseline compared to HCHO (p=0.049), the HCHO group improved by week 2 (p=0.009) and week 3 (p=0.007) whereas the LCHF group were significantly slower in Week 1 (p<0.001) and Week 2 (p=0.026) compared to Baseline. During the 14 km hill session, within-group analysis shows that athletes walked significantly faster (p=0.010) and at a higher percentage of vVO2max (p=0.007) when using caffeine compared to a placebo. Between-group differences remained present, with HCHO athletes walking at a higher percentage of vVO2max than those adhering to the LCHF diet (p=0.035). No interaction between supplement treatment and dietary group occurred (p=0.640). Caffeine supplementation partially reversed the performance impairment associated with an LCHF diet, but training quality remained lower than the combination of caffeine and high CHO availability.