The scientific community is becoming more demographically diverse, and team science is becoming more common. Here, we compare metrics of success in STEM, such as acceptance rates and citations, between differing team compositions regarding nationality, gender, career stage, and race/ethnicity. We collected the final decisions and citations as of 2019 of 91,427 manuscripts submitted from 2012-2018 to journals published by the American Geophysical Union. We matched the authors by email on each manuscript to self-provided demographic information within the American Geophysical Union’s membership database. This resulted in 20,940 manuscripts matched to nation, gender, and career stage, and 6,015 manuscripts matched to race/ethnicity for manuscripts whose entire authorship team was affiliated with the U.S. Among similar sized authorship teams (teams of 2-4), acceptance rates were 2.7, 4.5, and 0.9% higher (pnation < 0.01, pgender < 0.05, pcareer stage = 0.51) with more than one nation, gender, and career stage, respectively, than non-diverse authorship teams. Diverse papers had 1.2 more citations for international teams than single-nation teams (pnation < 0.01). There were 0.4 and 1.0 fewer citations for authorship teams with more than one gender or career stage than manuscripts with one gender or one career stage (pgender = 0.21, pcareer stage = 0.36). However, racially/ethnically diverse teams were associated with 5.5% lower acceptance rates (p < 0.01) and 0.8 fewer citations (p = 0.15) than racially/ethnically homogenous teams. These results show that diversity can have tangible benefits to science, but equitable practices and inclusive cultures must also be fostered.