Background: Students' beliefs about their ability to grow in STEM disciplines have been linked to better course outcomes. However, such mindset beliefs are subject to the environmental cues projected by the instructor in the classroom, which we refer to as the mindset context. Recent meta-analyses indicated heterogeneity in the benefits of student mindset interventions, which the classroom environment may shape. In this work, we use structural equation modeling (SEM) to investigate the mindset context and its impact on students' affect and performance in STEM courses, particularly for students from marginalized groups who may be disproportionately affected by these factors. Results: We collected student perceptions of their instructors' universality beliefs about student abilities (all people or only some people can reach excellence in STEM), students' growth beliefs, sense of belonging (as measured by peer support, faculty support, and classroom comfort) and course grades. The sample was collected from courses in a STEM college within a demographically diverse, moderately selective institution in the Southern United States (N = 625). We found that student perceptions of the mindset context did not directly predict course grades, but ACT scores did (standardized exams used for college entry in the USA). However, SEM analysis revealed that when students perceived instructors to believe only some students can succeed in STEM (endorse more non-universal beliefs), they reported fewer growth beliefs about their abilities in STEM. This led to less classroom comfort in contributing to class discussions, ultimately lowering STEM grades. Multigroup moderation analysis showed no differences in paths based on race, gender, and generational status. However, the mindset context impacted traditional students' (age of 18-22) growth beliefs to a greater extent than non-traditional students (> 22 years old). Additionally, classroom comfort significantly predicted grades for traditional students but not for non-traditional students. Conclusions: Our finding suggests that when students perceive the mindset context more positively, their outcomes improve, especially for traditional students who may be more sensitive to classroom cues. Thus, mindset interventions for faculty (coupled with student interventions) may also be beneficial to supporting student success. Additionally, we recommend improving student content preparation to enhance foundational knowledge, considering that indicators of prior preparation (ACT scores) play a more direct role in predicting student grades.