Hand posture estimation is a building block in hand gesture detection systems. One of the most popular techniques in this field is to generate a 3D hand model and find its optimal structure using an optimisation algorithm. The main advantages of such methods are flexibility to model complex hand postures, ability to better address occlusions, and estimating both discrete and contentious poses. However, rendering and finding optimal structure for a 3D hand model are computationally expensive. Popular models in the literature are made of mesh or simple components. This work first analyses a number of hand models in the literature. Then, a new hand model with simple components is proposed. The performance of the proposed model is benchmarked on 50 poses extracted from four standard datasets. The results show that the proposed 3D model shows superior results compared to the best models in the literature.