Computed tomography (CT)-derived skeletal muscle area (SMA) and skeletal muscle radiodensity (SMD) reflect distinctive quantitative and qualitative characteristics of skeletal muscles. However, data on whether CT-based muscle parameters, especially SMD, can predict muscle function is limited. In a prospective cohort, 1523 community-dwelling older adults who underwent abdominal CT scans and the countermovement two-legged jumping test on a ground reaction force platform were analyzed (mean age 74.7 years, 65.1% women). SMA and SMD were measured at third lumbar vertebra level (L3). Individuals with low jump power (peak weight-corrected jump power < 23.8 W/kg in men and < 19.0 W/kg in women using clinically validated threshold) were older;​ had lower SMA, SMD, and maximal grip strength values;​ and had lower chair rise test and timed up and go test performance than those without low jump power. SMD was positively associated with peak weight-corrected jump power (adjusted β = 0.33 and 0.23 per 1 HU increase in men and women, respectively, p < 0.001). One HU decrement in SMD was associated with 10% elevated odds of low jump power (adjusted OR [aOR] 1.10, p < 0.001) after adjusting for age, sex, height, inflammation, and insulin resistance markers, whereas the association of SMA with low jump power was attenuated (aOR 1.00, p = 0.721). SMD showed better discrimination for low jump power than SMA (AUC 0.699 vs. 0.617, p < 0.001), with additional improvement when added to SMA and conventional risk factors (AUC 0.745 to 0.773, p < 0.001). Therefore, CT-measured L3 SMD can be a sensitive surrogate marker for muscle function along with SMA in older adults, which merits further investigation.