Abstract: Traditional photodetectors exhibit rectification characteristics, resulting in high dark current, elevated noise levels, and reduced sensitivity under forward bias. Consequently, these devices must operate in reverse bias mode based on the photovoltaic effect, commonly referred to as PV mode. However, accurate data acquisition remains challenging for weak optical signals due to the limited performance in this mode. Moreover, since the external quantum efficiency (EQE) is below 100%, additional readout circuits are required to amplify weak signal variations. Therefore, photomultiplier-type photodetectors offer significant potential for weak-light detection and related applications. This paper reports a dual-mode organic photodetector (OPD) that integrates both photoelectric multiplication (PM) and synaptic plasticity, along with its application in handwritten digit recognition. The device achieves a photocurrent amplification factor of 10² under forward bias via the photoelectric multiplication effect and approaches a factor of 10³ under reverse bias in photovoltaic mode. Using the MNIST dataset for handwritten digit recognition, a pattern recognition accuracy of 98.36% is attained.