Watch: Chain of Thought
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.
Transcript
When a language model works through a problem step by step before giving a final answer, we call it a chain of thought. Instead of jumping straight to a conclusion, the model lays out its reasoning, calculations, or plans. This simple shift dramatically improves accuracy on complex tasks like math, logic, and deep analysis.
It all started as a prompting technique. By simply asking a model to think step by step, users prompted it to generate intermediate steps, which naturally led to a better final answer.
Today, reasoning models take this a step further. They are trained to produce a long, internal chain of thought automatically, spending more computing power on harder problems. This deliberation is built right into the model. Even though this intermediate reasoning is often hidden from the user or shown only as a brief summary, it actively shapes the final output.
For information visibility, this shift is crucial. Because these models run multiple sub-queries and weigh various sources before settling on an answer, being discoverable across the many angles the model explores is far more important than just matching a single search phrase.