BACKGROUND Osteoarthritis is a common degenerative disease of joints, and animal models have important significance in the investigation of this disease. The aim of this study was to develop a better method for developing osteoarthritis models in primates by comparing the modified Hulth score combined with joint scratches modeling method with others. MATERIAL AND METHODS We randomly divided 15 young male Macaca fascicularis and 3 old male Macaca fascicularis into 6 groups (n=3). Knee osteoarthritis (KOA) models were developed with different methods: modified Hulth combined with joint scratches (Group A), modified Hulth (Group B), Hulth (Group C), spontaneous models (Group D); sham-operated (Group E), and blank control (Group F). Morphology and pathology of knee joints were observed at the 8th week after surgery. The levels of WBC, IL-1b, and TGF-b1 in synovial fluid were detected by ELISA. The levels of COL-II, ACAN, and MMP-13 in articular cartilage were examined by RT-qPCR and Western blot. RESULTS In Brittberg and modified Mankin score, Group A was higher than B (P<0.05) and lower than C (P<0.05), and there was no statistically significant difference between Group A and D (P>0.05). Except for Group E and F, the differences were statistically significant among others in WBC, IL-1β, and TGF-β1 (P<0.05). COL-II and ACAN decreased and MMP-13 increased, and there was no significant difference between Groups A and D (P>0.05) or between Groups E and F (P>0.05). There were statistically significant differences among other groups (P<0.05). CONCLUSIONS The models developed by modified Hulth combined with joint scratches were the closet to spontaneous models at the 8th week after surgery.