Objective To compare and analyze the differences between two commonly used modeling method for rabbit knee osteoarthritis (KOA) and provide references for the selection and establishment of different types of animal models for knee osteoarthritis. Methods Nine healthy New Zealand white rabbits were randomly divided into 3 groups: blank group, modified Hulth group and collagenase group. The model groups were treated with a modified Hulth method or intra-articular injection of collagenase. After 1 week, the rabbits were encouraged to move daily. After 6 weeks, they were evaluated using Lequesne MG scores, X-rays, micro-computed tomography, Pelletier scores, Mankin scores and enzyme- linked immunosorbent assays, and then the gross and pathological changes in the knee were observed. Results Compared with that in the blank group, the cartilage surface of the rabbit knee joints in the two model groups displayed defects. The pathological result showed that the cartilage layer appeared to be cracked, safranin O staining was decreased, the tide line was distorted, and the blood vessels passed. The concentrations of TNF-α in serum from the blank group, modified Hulth group, and collagenase group were (624. 99 ± 17. 82), (1140. 56 ± 129. 81) and (1480. 69 ± 492. 08) pg / mL, and the IL-1β concentrations in serum were (30. 23 ± 0. 25), (46. 67 ± 0. 71) and (46. 82 ± 1. 04) pg / mL, respectively. The modified Hulth and collagenase groups showed statistically significant differences compared with that in the blank group (all P<0. 05). In addition, the modified Hulth group was accompanied by osteophyte formation in the knee joint and more serious damage in the medial femoral condyle cartilage, while the medial tibial condyle was the main injury in the collagenase group. Conclusions Both modeling method can be used to construct an ideal KOA model. The degree of damage using the modified Hulth method was more serious, while the collagenase-induced model was closer to the process and pathogenesis of human KOA.