Term explained
What is a reasoning model?
A reasoning model is an AI model that works through a problem step by step before answering, which makes it better at maths, coding and planning but slower.
In plain English
A standard chatbot starts writing its answer straight away, word by word. A reasoning model first “thinks”: it writes out a hidden chain of working, tries an approach, checks it, and backs up if it spots a mistake, much like doing a sum on scrap paper before writing the answer down. Only then does it give you its reply.
The approach went mainstream in September 2024, when OpenAI released its o1 model. Today most major AI companies offer reasoning, often as a “thinking” mode or a setting for how hard the model should think.
Why it matters
Thinking before answering makes a big difference on problems with several steps: maths, coding, analysing a contract, or planning a trip with lots of constraints. Reasoning models score far higher on hard benchmarks than models that answer instantly, and they power many AI agents that carry out long tasks.
An example
Ask a quick model how many Mondays there are between two dates and it may guess. A reasoning model will work out what day the first date falls on, count the weeks, check the leftover days and then answer, and it is much more likely to be right.
What to watch out for
Reasoning takes time and costs more, because all that hidden working counts as extra tokens. For simple questions, a quick answer is fine. Reasoning models also still make mistakes, and their step-by-step summaries can sound more convincing than they deserve, so check the answers that matter.