This study investigates the diagnostic accuracy of the combination of usual walking speed (UWS) and maximum walking speed (MWS) to identify frailty in community-dwelling older adults. A population-based study with 758 participants aged 65 and older was conducted. Frailty syndrome was determined using the Fried phenotype. UWS and MWS were evaluated in a 4.6-meter path. Both measures were categorized using the 1.0 m/s cut points, and participants were categorized into three groups: those with “very good”, “good” and “insufficient” walking reserve capacity (WRC). Of all participants, 9% were identified as frail and 47% as prefrail. The “insufficient” WRC presented a low sensitivity of 0.55, high specificity of 0.91 and moderately useful likelihood ratios (LR+ 6.57, LR- 0.48) to identify frailty. Based on Fagan’s nomogram, an elder’s corresponding post-test probability of being frail with an “insufficient” WRC would be around 40%, which substantially increased the diagnostic accuracy of frailty.