Artículo de publicación ISIThis study examined the differences in the bench press one-repetition maximum obtained by three different methods (direct method, lifts-to-failure method, and two-point method). Twenty young men were tested in four different sessions. A single grip width (close, medium, wide, or self-selected) was randomly used on each session. Each session consisted of an incremental loading test until reaching the one-repetition maximum, followed by a single set of lifts-to-failure against the 75% one-repetition maximum load. The last load lifted during the incremental loading test was considered the actual one-repetition maximum (direct method). The one-repetition maximum was also predicted using the Mayhew’s equation (lifts-to-failure method) and the individual load–velocity relationship modeled from two data points (two-point method). The actual one-repetition maximum was underestimated by the lifts-to-failure method (range: 1–2 kg) and overestimated by the two-point method (range: –3 to –1 kg), being these differences accentuated using closer grip widths. All predicted one-repetition maximums were practically perfectly correlated with the actual one-repetition maximum (r ≥ 0.95; standard error of the estimate ≤ 4 kg). The one-repetition maximum was higher using the medium grip width (83 ± 3 kg) compared to the close (80 ± 3 kg) and wide (79 ± 3 kg) grip widths (P ≤ 0.025), while no significant differences were observed between the medium and self-selected (81 ± 3 kg) grip widths (P = 1.000). In conclusion, although both the Mayhew’s equation and the two-point method are able to predict the actual one-repetition maximum with an acceptable precision, the differences between the actual and predicted one-repetition maximums seem to increase when using close grip widths.