Haematogenous models of septic arthritis have some inherent disadvantages, such as the manifestation of arthritis relies on chance, the size of the inoculum is unknown and the number of animals to be studied cannot be reduced because the animals cannot serve as their own controls. This study aimed to develop a rat model of knee septic arthritis by injecting a known inoculum of Staphylococcus aureus into the joint. The left knees of 27 Sprague Dawley rats were injected with four different inoculum concentrations of a sensitive strain of S. aureus (30,000 colony-forming units (CFUs), n = 3; 18,550 CFUs, n = 6; 15,500 CFUs, n = 9; and 7700 CFUs, n = 9); the right knees served as controls. Clinical, microbiological and histological variables were assessed two and seven days later. The main criterion for diagnosing septic arthritis was a positive culture of synovial fluid. The rate of microbiologically confirmed septic arthritis was high for all inoculum concentrations (3/3, 6/6, 8/9 and 7/9, respectively), and the rate of bacteraemia was also high. Animal welfare was better for the two lowest inoculum concentrations. No animal reached the pre-established humane end points. Overall, the third inoculum was considered the most suitable. Thus, acute septic arthritis can be caused in rats by inoculating 15,000 CFUs of an ATCC strain of S. aureus directly into the knee joint. Overall, the model seems to be useful for studying the effectiveness of drugs for the treatment of acute septic arthritis.