Birth at term comprises a period with heterogeneous neonatal outcomes that tend to be worse for infants born earlier. However, few studies have analyzed this period, in which each day can make a difference. Therefore, we aim to assess neonatal mortality (NM) according to gestational age (GA) at birth measured in days in term liveborn infants born in 2012-2017 in São Paulo, the largest city in Latin America. This population-based cohort study assessed term liveborn infants followed until the end of the neonatal period. We analyzed 7 models for NM according to GA in days: crude NM adjusted for maternal and prenatal variables, NM additionally adjusted for type of birth and type of hospital, and adjusted NM stratified by type of birth (cesarean and vaginal) and by type of hospital (public and private). We included 440,119 live infants born at 259-293 days of gestation. The median GA at birth was 274 days. In all models, NM was higher for infants born early term, decreasing in infants born full term and rising again in infants born late term. In the unadjusted model, hazard ratios of NM changed daily, decreasing from 3.34 to 1.00 on day 278 and increasing again thereafter. In the stratified analysis according to type of hospital, being born in a public hospital was associated with a reduced risk of NM for infants born at 278-283 days of pregnancy. There was a decrease in GA related to obstetric interventions, especially cesarean sections, which increased NM. The loss of days of pregnancy was larger in private hospitals. Increasing the granularity of GA to days is feasible and has the potential to drive public policies. To the best of our knowledge, this is the first Brazilian study on GA in days using a national live births database.