In horticulture, different planting designs can be used to optimize the plant production. In the coffee tree, we assumed that different planting densities and spatial patterns may impact the growth of all branching order axes, which in turn would impact the berry production. The aim of this study was to quantify and compare axes growth, leaf area, berry distribution and yield in four production years, depending on four planting designs. Experimental data were collected in the 1st, 2nd, 6th and 7th production years (PY), on Coffea arabica trees planted in two high densities (6000 or 10,000 plants ha⁻¹) and planting patterns (square or rectangular). Coffee architecture was described from metamer to axis, layer, and plant scales. Metamer number per axis, axes number and cumulated length, together with berry distribution, were compared along 40 cm-thick layers of plant vertical profile, depending on the planting design. In the 1st and 2nd PY, plants were formed by two layers, with axes of 4th order appearing in layer 1. In the 2nd PY, the 3rd order axes which had a crucial role in total berry production, were about 3–4 folds more numerous than 2nd order axes in layer 1. In the 6th and 7th PY, the plants included five layers and five branching orders. The leaf area index, 2nd – 4th order axes length, berry distribution and yield were shown to depend on both planting design and PY. Both planting patterns under 10,000 plants ha^-1 and rectangular planting pattern under 6000 plants ha⁻¹ could be recommended for high production up to 7th PY, while the square planting pattern under 6000 plants ha⁻¹ must be pruned after 6th PY.