This case study aims to investigate the relationship between the online reading group meetings (ORG) and reading difficulties among upper-level learners of Chinese as a foreign language (CFL learners). Based on Bernhardt’s constructivist model of L2 reading (1986, 1991) and sociocultural perspectives, this study attempts to answer two research questions: 1. What reading difficulties upper-level CFL learners encountered when they read in Chinese? 2. What components in the online reading group meetings helped upper-level CFL learners to cope with their reading difficulties? How did the components help? Seven Fourth-Year Chinese learners at a university in the United States participated in this study and attended the ORG meetings throughout a semester to discuss five assigned readings. Qualitative research methods were used to collect and analyze the data. The data sources of this study included pre- and post-ORG questionnaires, pre- and post-meeting comprehension checks, pre- and post-meeting self-reports, recordings of the online reading group meetings, and semi-structured interviews. The findings showed a complicated and interactive relationship of the six components in Bernhardt’s model when it came to reading difficulties among upper-level CFL learners. Even though the participants identified word recognition as one of their top reading difficulties when reading in Chinese, the findings showed that extra-text driven components, especially intratextual perceptions and prior knowledge, played an important role in determining the participants’ reading comprehension. Being able to constantly monitor what they learned from each paragraph and which prior knowledge they applied to the reading was found to be a key to improving their reading comprehension. The findings also found reflexivity in dialogic collaboration, especially when the ORG meetings created opportunities for the participants to improve their reading comprehension via peer-to-peer interaction. The patterns of peer-mediated learning included giving corrective feedback, negotiating meaning, learning from listening, strategy coaching, and using English strategically. Whereas the instructor as a facilitator was regarded important to provide timely feedback and facilitate the discussion, the findings suggested that a peer-led group discussion was possible for upper-level CFL learners once they built up their confidence in offering help to their peers. Even though the component of the role assignment seemed to be beneficial for the participants to cope with their reading difficulties in the ORG meetings, the participants did not perceive each role equally helpful nor easy due to various reasons. The study concluded with a discussion of the implications for CFL education, teacher education programs and future research.