BTC
ETH
SOL
XRP
BNB
TON
DOT
LTC
POL
HBAR
NEAR
ARB
UNI
An AI model that works through a problem step by step before answering, which helps with maths, code and planning but takes longer.
AI Drafted with AI help · Reviewed by Keshav Aggarwal · Updated
In simple words
Some students blurt out the first answer. Others write their working on rough paper first, then give the answer.
A reasoning model is the second kind of student.
The technical side
Reasoning models are trained, often with reinforcement learning, to produce a long internal chain of thought before the final answer. OpenAI’s o1 (September 2024) brought the approach to the mainstream, and DeepSeek-R1 (January 2025) showed an open-weights version.
They score higher on hard maths, science and coding tests, but use many more output tokens, so they are slower and cost more per answer. Compare prices on our LLM API prices page.
Reasoning Model kya hai?
Reasoning model jawab dene se pehle step by step sochta hai, jaise rough kaam karke sawaal hal karna. Dheema hai par mushkil sawaalon mein behtar.
Related terms
- LLM (Large Language Model)AIAn AI model trained on huge amounts of text so it can read and write language, like the models behind ChatGPT, Claude and Gemini.
- InferenceAIThe moment an AI model is used to answer: it takes your input and works out the output. Each answer you get is one inference.
- HallucinationAIWhen an AI model states something false or made up as if it were true, such as a fake fact, quote or source.