AIM: The role of environmental conditions in limiting species distributions is often hypothesized, but it is challenging to gather large-scale data to demonstrate environmental impacts on individual performance. The past and present biogeography of model organisms is key context to understanding how environment shapes species' genetic and phenotypic diversity. LOCATION: Global. TAXON: Arabidopsis thaliana ("Arabidopsis"). METHODS: We fit occurrence records to climate data, and then projected the distribution of Arabidopsis under last glacial maximum, current, and future climates. We confronted model predictions with individual performance measured on 2,194 herbarium specimens, and we asked whether predicted suitability was associated with life-history variation measured on 898 natural accessions. RESULTS: The most important climate variables constraining the Arabidopsis distribution were winter cold in northern and high elevation regions and summer heat in southern regions. Herbarium specimens from regions with lower habitat suitability in both northern and southern regions were smaller, supporting the hypothesis that the distribution of Arabidopsis is constrained by climate-associated factors. Climate anomalies partly explained interannual variation in herbarium specimen size, but these did not closely correspond to local limiting factors identified in the distribution model. Late-flowering genotypes were absent from the lowest suitability regions, suggesting slower life histories are only viable closer to the center of the realized niche. We identified glacial refugia farther north than previously recognized, as well as refugia concordant with previous population genetic findings. Lower latitude populations, known to be genetically distinct, are most threatened by future climate change. The recently colonized range of Arabidopsis was well-predicted by our native-range model applied to certain regions but not others, suggesting it has colonized novel climates. MAIN CONCLUSIONS: Integration of distribution models with performance data from vast natural history collections is a route forward for testing hypotheses about species distributions and their relationship with evolutionary fitness across large scales.