A compact, openly available model from the tomg-group-umd research group. At 1.7B parameters with a 40K token context window, it punches above its weight class in terms of context handling relative to its size. Details about its specific capabilities and intended use cases are limited beyond its text-in, text-out design.