Artículo de publicación ISIObjective: To compare the accuracy of different devices to predict the bench-press 1-repetition maximum (1RM) from the individual load-velocity relationship modeled through the multiple- and 2-point methods. Methods: Eleven men performed an incremental test on a Smith machine against 5 loads (45-55-65-75-85% 1RM), followed by 1RM attempts. The mean velocity was simultaneously measured by 1 linear velocity transducer (T-Force), 2 linear position transducers (Chronojump and Speed4Lift), 1 camera-based optoelectronic system (Velowin), 2 inertial measurement units (PUSH Band and Beast Sensor), and 1 smartphone application (My Lift). The velocity recorded at the 5 loads (45-55-65-75-85% 1RM), or only at the 2 most distant loads (45-85%1RM), was considered for the multiple- and 2-point methods, respectively. Results: An acceptable and comparable accuracy in the estimation of the 1RM was observed for the T-Force, Chronojump, Speed4Lift, Velowin, and My Lift when using both the multiple- and 2-point methods (effect size < 0.40; Pearson correlation coefficient [r] > .94; standard error of the estimate [SEE] <4.46 kg), whereas the accuracy of the PUSH (effect size = 0.70-0.83; r = . 9 3 -9 4 ; SEE = 4.45- 4.80 kg), and especially the Beast Sensor (effect size = 0.36-0.84; r = . 50-68; SEE = 9.44-11.2 kg), was lower. Conclusions: These results highlight that the accuracy of 1RM prediction methods based on movement velocity is device dependent, with the inertial measurement units providing the least accurate estimate of the 1RM.