A Theory of Magnitude (ATOM) proposes that time, quantity and space share an innate and common metric system. An increasing number of empirical works have demonstrated that our time perception is influenced by other concurrent magnitude attributes such as space, number, weight and luminance. However, it is still unknown that when and where the numerical magnitude effects in time perception occurs. Our research used duration reproduction task and numerical processing task synchronously to explore how numerical processing affects temporal reproduction. In temporal reproduction task, a dot, as standard stimulus, appeared on screen for one of the standard durations(300ms, 450ms, 600ms, 750ms). Participants then continuously pressed the space bar to reproduce the duration of the standard stimulus. While participants pressing the space bar, a digit(1, 2, 8, 9) as reproduction stimulus, was presented on screen . In Experiment 1, 25 participants were required to name the digit. In Experiment 2, 29 participants were required to determine the parity of the digit. In Experiment 3, 30 participants were required to determine whether the digit was smaller or larger than 5. We observed the numerical magnitude effects in Experiment 1 and Experiment 3, but not in Experiment 2. In Experiment 1 participants were required to name the digit. The effects of numerical magnitude were different on different duration reproductions. Reproduced durations were longer for big numbers and shorter for small numbers in the 300ms standard duration. However, the effects disappeared in the 450ms standard duration. In contrast, reproduced durations were shorter for big numbers in the 600ms standard duration. In the 750ms standard duration, the effects disappeared again. In Experiment 3 participants were required to determine whether the digit was smaller or larger than 5. Similar with the result in Experiment 1, Experiment 3 revealed that under different standard durations the effects of numerical magnitude on temporal reproduction were different. Participants’ reproduced durations in the 300ms standard duration were longer for big numbers and shorter for small numbers. The reproduced durations in the 450ms standard duration were shorter for big numbers. The effects disappeared in the reproductions of 600ms standard duration. The reproduced durations in the 750ms standard duration were longer again for big numbers. All in all, our results show that the temporal perception was influenced by numerical magnitude in both the numerical naming task and numerical judging task. Moreover, the effects of numerical magnitude might be different in different standard durations and numerical processing tasks. The findings suggest that the numerical magnitude effects in time perception change dynamically according to numerical processing conditions and the effects could occur at any stage of temporal cognitive processes.