Background: Smoking during pregnancy has been declining in the past decades in high-income countries, including Sweden. Paradoxically, increasing trends associated with duration of residence have been reported among immigrants. We aimed to clarify how these two contrasting trends have shaped smoking patterns among immigrants. Methods: We conducted a population-based study of 1 598 433 pregnancies in Sweden in the period 1992–2008. We used multinomial logistic regression to estimate the odds of mild and heavy smoking relative to no smoking associated with year of delivery, years since migration, maternal region of birth and their interaction, after controlling for potential confounders. Results: The prevalence of smoking decreased for the Swedish-born and for immigrants during the study period. Among immigrants, duration of residence was independently associated with increases in smoking and varied according to maternal region of birth (P-value for interaction <0.001). The odds ratio associated with a 10-year increase in duration of residence was weakest for mild smoking among former Yugoslav women (adjusted odds ratio; 95% confidence interval: 1.10; 1.04–1.17) and those from other Nordic countries (1.22; 1.17–1.26) and strongest for heavy smoking among East African (4.46; 3.23–6.16) and sub-Saharan African (3.56; 2.68–4.72) women. The association between duration of residence and smoking was attenuated after stratifying by cohorts of arrival among European but not among non-European immigrants. Conclusions: Declines in smoking during pregnancy among immigrants from various regions of the world were differentially affected by opposite increasing trends throughout their residence in Sweden.