There is a well-known deficiency in valid animal models for bipolar disorder. Developing the single ideal model for the disorder-one that will represent its full scope-will probably not be possible until we have a much better understanding of the underlying pathology. Yet, intermediate models, even with partial validity, are critical in order to advance our knowledge and put us into position to develop even better models. The present article discusses the various efforts under way to develop the best models based on our current level of understanding. These efforts include (1) identifying new tests, (2) developing models based on the endophenotypes approach, (3) identifying the best rodent strains, (4) identifying the most appropriate species, (5) segregating susceptible versus resilient animals, and (6) segregating animals that respond or do not respond to treatment. It is suggested that a combined approach that includes these directions and others can result in better models with higher validity that will offer significant help in advancing research on bipolar disorder and developing new and better treatments.