Abstract Identifying and quantifying factors that influence tree growth are crucial issues to ensure sustainable forest management, particularly in moist tropical forests. Tree growth depends on several factors comprising ontogenic stage, competition by neighbours and environmental conditions. Several studies have focused on one or two of them, but very few have considered all three, especially in Central Africa. We investigated the effects of diameter and competition on tree growth, in four Central African sites characterized by their soil physicochemical properties, at both tree community and population levels. We calibrated growth models using diameter data collected on 29,741 trees between 2015 and 2018, on twelve 4 or 9‐ha plots spread over the four sites. These models included diameter, wood density, competition indices and site effect as explainable variables at the community level and excluded wood density at the population level. At the community level, the best models explained 11% of growth variability with a decreasing effect of species wood density, diameter, site and competition. Our results show that even if low, site effect can result from different soil nutrients depending on both tree size and species wood density. We observed higher tree growth on sites with (i) high exchangeable K, organic C, total N and total P for low wood density species; (ii) high available P and C:N for small trees, high exchangeable Ca and Mg for medium to large trees, all belonging to medium and hard wood density species. At the population level, the best models explained between 0 to 43% of growth variability, with significant competition effect (resp. site effect) for 21 (resp. 9) of the 43 species studied. Site ranking varied greatly between the 9 species concerned, probably reflecting different sensitivities to the scarcity of particular soil nutrients. Synthesis. Our study provides original results on the factors influencing tree growth in Central Africa, showing that the potential effect of soil nutrients depends on tree size and species wood density. Remaining highly unpredictable at the population level, this effect makes it essential to increase the number of dynamics monitoring systems in logging concessions.