Spoken sentence processing often requires integrating the linguistic message with information about the talker and the social context. Among other things, information about the talker’s gender identity could influence how a listener processes what they hear, due to the prevalence of gender-related stereotypes in human societies. Several previous studies showed that listeners anticipated stereotype-consistent content, and that comprehension was affected when gender stereotypes were violated (e.g., when women serve in conventionally male-dominant professions, or vice versa). However, the existing findings are rather mixed. In this study, we examine the influence of talker gender information on language comprehension and prediction in Mandarin Chinese. We report the results of a cloze task where participants were asked to guess the last word of a sentence with or without information about talker gender. When talker gender information was available, we further varied whether gender information was revealed by personal names or voices. Participant responses were evaluated for gender bias by two pre-trained language models based on Word2Vec and GPT2. Results of statistical analysis revealed that participants adjusted their responses to align with the gender category cued by the sentence (i.e., more male/female-biased responses when the sentences implicated a male/female talker), but the effect was only present when gender information was implicated through names but not voices. The current study provides partial evidence for the effect of talker gender on sentence prediction. We discuss the implications of current findings for further study of the integration of linguistic and social information in language comprehension.