The acquisition of voicing contrast in French stops is one of the most significant challenges for Mandarin speakers learning French, due to the absence of voiced categories in their native language. Nevertheless, Wu distinguishes itself from other Sinitic languages in having a three- way contrast between voiceless aspirated, voiceless unaspirated and voiced stops. The voiced series in the initial position of words, however, becomes phonetically voiceless while incorpo- rating breathy phonation (Chao, 1956; Cao & Maddieson, 1992). According to the Perceptual Assimilation Model (PAM) (Best, 1995; Best & Tyler, 2007), cross-linguistic perception is in- fluenced by L1 phonological knowledge. Conversely, the Speech Learning Model (SLM) (Flege, 1995) postulates that performance in second language speech learning is contingent upon the surface phonetic similarities and dissimilarities between L1 and L2 sounds. Furthermore, SLM suggests that more L2 experience enhances learners’ ability to perceive and produce L2 sounds more effectively. From these predictions arise the following questions: 1) Can Wu speakers demonstrate a more proficient acquisition of the voicing contrast in French compared to the Mandarin group, given that voiced and voiceless plosives are classified into distinct categories in their L1? 2) Is the voicing contrast discerned more easily by the Wu group in word-medial position, where phonologically voiced initial stops are truly voiced? 3) Does a longer period of French learning lead to a better perception and production? Considering the limited number of research studies that compared the performance of Wu and non-Wu speakers in acquiring French stops (Qin, 2010; Wang, 2020; Jin & Chabanal, 2022), the present paper conducted two types of tasks: a categorizing task and a reading task including 30 word pairs. This involved 30 participants who were native speakers of Wu and Mandarin, each with different proficiency levels in French. After measuring the accuracy rate and several acoustic cues using Praat and VoiceSauce softwares, we applied a series of multilevel regression models in the R environment to examine the influence of four predictors, namely L1, consonant, syllable position and duration of learning French. The results show that Wu speakers exhibit a high proficiency level in perceiving and producing the voicing contrast of French stops. In particular, the voiced categories produced by the Wu group are characterized by a more nega- tive VOT and a greater V-ratio. We also observed that some Wu and Mandarin speakers use the following strategy to differentiate the French voicing contrast: nasalization of voiced stops in initial position and aspirated production of their voiceless counterparts. Moreover, the in- tervocalic position does not facilitate the acquisition of voicing distinction among Wu speakers. In fact, the phonetic cues used by these learners are VOT and V-ratio instead of phonation differences (f0, H1-H2, H1-A1, H1-A2 and CPP). In addition, an extended duration of French language learning does not appear to correlate with an enhanced perceptual capacity but rather with improved production. Together, this study illustrates that both the PAM and the SLM achieve only partial success in explaining our results.