AI glossary
Reasoning model
A language model trained to think through a problem internally before answering, trading extra time and tokens for better results on hard tasks.
Reasoning models generate a hidden or summarised chain of thought: they plan, try approaches, check their work and only then reply. This helps with maths, coding, analysis and multi-step agent tasks.
Most current flagship models reason by default and let you control how much through an effort setting. More effort costs more tokens and time, so simple questions are better served with low effort.
Example: Faced with a logic puzzle with several conditions, a reasoning model tries a hypothesis, sees it contradicts one of the clues, drops it and tries another before answering. It takes longer but gets it right more often.
In practice
- Use them for multi-step problems: analysis, maths, code or planning.
- For simple questions, lower the effort level or use a fast model.
- Reasoning is billed as output tokens even when you do not see it.
Which models reason and what they cost: AI model guide.
