A model working through a problem in explicit intermediate steps before answering; reasoning models build this deliberation in rather than relying on the prompt to elicit it.
Chain of Thought is a language model working through a problem in explicit intermediate steps before giving a final answer, rather than jumping straight to it. Laying out the reasoning, a calculation, a deduction, a plan, markedly improves accuracy on multi-step tasks in maths, logic, and analysis.
It began as a prompting technique: asking a model to "think step by step" makes it generate those steps, and the steps condition a better answer, related to how generative self-retrieval lets a model surface facts by writing them out. Reasoning models take this further. Trained to produce a long internal chain of thought before answering, and to spend more inference compute on harder problems, they build the deliberation into the model rather than relying on the prompt to elicit it.
The intermediate reasoning is often hidden from the user, shown as a summary or withheld entirely, while it still shapes the output. For AI visibility, more deliberate reasoning means a model can run several sub-queries and weigh sources before it settles an answer, so being retrievable and citable across the many angles a reasoning model explores matters more than matching a single phrasing.
