# Reasoning Model

> A model that thinks step by step before it answers. Slower, but better at hard problems.

**Field:** Artificial Intelligence  
**Category:** LLMs  
**Also known as:** Thinking Model, Reasoning LLM

## Think of it like

Think of it like a kid who works the sum out on scrap paper first. Then they write the answer.

## Definition

A reasoning model is a language model trained to work through a problem step by step before it gives an answer. It writes out a chain of thinking, checks its steps and can change course, then gives a final reply. This makes it much better at maths, code and puzzles. The thinking costs extra tokens, so answers are slower and cost more. The first widely used ones came out in late 2024.

## Example

Lucy asks a hard logic puzzle. A normal chat model answers in 2 seconds and gets it wrong. A reasoning model thinks for 40 seconds, then gets it right.

## Why it matters

For maths, code and planning, the time spent thinking buys real gains in accuracy.

## Common mistake

Using it for every task. For a quick rewrite or a simple fact, the extra thinking only adds cost and wait.

## Three principles

- **Thinks first.** Step by step.
- **Better on hard tasks.** Maths, code, puzzles.
- **Costs more.** Extra tokens and time.

## Related terms

- [Instruct Model](https://learn.benibix.com/term/ai/instruct-model)
- [Large Language Model (LLM)](https://learn.benibix.com/term/ai/large-language-model)
- [Next-Token Prediction](https://learn.benibix.com/term/ai/next-token-prediction)
- [Chain-of-Thought Prompting](https://learn.benibix.com/term/ai/chain-of-thought-prompting)
- [Extended Thinking](https://learn.benibix.com/term/ai/extended-thinking)
- [Test-Time Compute](https://learn.benibix.com/term/ai/test-time-compute)
- [RL with Verifiable Rewards (RLVR)](https://learn.benibix.com/term/ai/reinforcement-learning-with-verifiable-rewards)

## Told apart from

- [Reasoning Model vs Instruct Model](https://learn.benibix.com/compare/reasoning-model-vs-instruct-model): People often think a reasoning model is just a smarter chat model. It is trained to think out loud first, spending extra tokens on steps before it answers. An instruct model answers straight away. For hard maths or code the thinking pays off; for quick tasks it only adds time and cost.

## Taught in

- [Reasoning Models and Thinking](https://learn.benibix.com/learn/ai-llms/reasoning-models): LLMs, 12/15
- [Chain-of-Thought Prompting](https://learn.benibix.com/learn/ai-working-with-ai/chain-of-thought-prompting): Prompt Engineering, 7/16

![Two paths from a question to an answer. The top path goes straight to an answer. The bottom path first passes through a box of thinking steps, then gives the answer.](https://learn.benibix.com/diagrams/llm-reasoning.svg)

---

Source: https://learn.benibix.com/term/ai/reasoning-model
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-18
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
