OBJECTIVE: To determine mortality and retention in ART programmes in Tanzania, between 2010 and 2013. METHODS: Retrospective routinely collected data were analysed from people starting ART in the period 2010-2013. Mortality and retention over the first 12 months on ART were compared across the 4 years, and adjustment was made for individual level potential confounders. RESULTS: Data from 3844 people (70.6% female) starting ART were analysed. Mortality in the first year declined from 11.4% in 2010 to 4.9% in 2013, and retention after 12 months increased from 77.8% in 2010 to 98.1% in 2013. Mortality was inversely associated with CD4 count, lowest among those with CD4 350+ cells/μl [adjusted odds ratio (AOR) = 0.03, 95% CI 0.01-0.03], associated with male sex (AOR = 1.79, 95% CI 1.39-2.31), but not age. Lost to follow-up (LTFU) was lowest among those with CD4 = 350+ cells/μl AOR = 0.20, 95% CI 0.10-0.30), but not associated with age or sex, and higher in urban health facilities (AOR = 1.88, 95% CI 1.15-3.09). After adjustment for individual level characteristics, there was a statistically significant yearly improvement in mortality (AOR = 0.31, 95% CI (0.21-0.44) and LTFU (AOR = 0.06, 95% CI 0.04-0.10). CONCLUSION: Mortality and retention in the first 12 months on ART have significantly improved over the 4 years from 2010 to 2013. These improvements may indicate better services, earlier initiation on ART, and strengthened systems to provide ART in Tanzania. These results refute the worries that earlier initiation on ART might lead to lower retention in the ART programme.