Reasoning models are advanced large language models trained, often using reinforcement learning, to perform complex reasoning tasks by generating structured chains of thought before producing answers. They are particularly effective in areas like problem-solving, programming, scientific reasoning, and multi-step planning.