Animal models of arthritis are used to evaluate potential antiarthritis drugs for clinical use. Therefore capacity of the model to predict efficacy in human disease is one of the most important criteria in model selection. Animal models of rheumatoid arthritis (RA) with a proven track record of predictability include rat adjuvant arthritis, rat type II collagen arthritis, mouse type II collagen arthritis, and antigen-induced arthritis in several species. Agents currently in clinical use (or trials) that are active in these models include corticosteroids, methotrexate, nonsteroidal anti-inflammatory drugs, cyclosporin A, leflunomide, interleukin-1 receptor antagonist, and soluble tumor necrosis factor receptors. For some of these agents, the models also predict that toxicities seen at higher doses for prolonged periods would preclude dosing in humans at levels that might provide disease-modifying effects. Animal models of osteoarthritis (OA) include mouse and guinea pig spontaneous OA, meniscectomy and ligament transection in guinea pigs, meniscectomy in rabbits, and meniscectomy and cruciate transection in dogs. None of these models have a proven track record of predictability in human disease because there are no agents that have been proven to provide anything other than symptomatic relief in human OA. Efficacy data and features of the various models of RA and OA are discussed with emphasis on their proven relevance to human disease.