This study shows that two distinct social models predict speech rhythm variation -measured by the normalized pairwise variability index of vowels (nPVI-V) -for Stockholm's two working classes. The non-white working-class variety (multiethnolect) has less intervocalic durational contrast than the speech of elites (41-51 vs. 49-57) and correlates with the speaker's neighborhood diversity. Incremental increases in neighborhood diversity correlate with incremental decreases in nPVI-V. The white working-class variety has more intervocalic durational contrast than the speech of elites (53-61 vs. 49-57) and correlates to occupational status. Incremental decreases in occupational status correlate with incremental increases in nPVI-V.The data comes from 31 male Stockholmers, ages 24-49, who read aloud a passage with 285 vocalic elements. Thirteen self-identify as white 'Swedes': five working class, eight uppermiddle class ('elites'). Eighteen self-identify as non-white 'immigrants': five working class, seven lower-middle class, six upper-middle class ('elites'). Twenty-eight were born in Sweden; three arrived before age four. They hail from five neighborhood types that are representative of Stockholm's geographic ethnic distributions.The findings add Swedish multiethnolect to a growing list of contact varieties with less intervocalic durational contrast than their heritage counterparts. The findings also nudge our field's discussion of rhythm away from second-language acquisition to the social domain of race and class. At the same time, a new research question emerges whether intervocalic durational contrast is a sociolinguistic variable in its own right or a byproduct of segment-level variation.