System One Model

Drawing on the dual-process theory from psychology, this term refers to lightweight AI models trained for specific narrow tasks. They operate with very low latency and cost within a constrained decision space, forming a complementary tiered architecture alongside general-purpose large models that require complex reasoning, referred to as System Two.