AI 101

Reasoning

Also known as: reasoning ability · logical reasoning

Inference

Reasoning is a model’s ability to work through multi-step logic — chaining facts, applying rules, checking intermediate results — rather than just pattern-matching an answer.

It’s the skill that lets models solve math, plan, and debug.

  • Weak reasoning: guesses the likely answer.
  • Strong reasoning: works it out step by step and can explain each step.

How to test it: give a problem with a known trap and see if the model catches it, or just produces the plausible-but-wrong answer. That's reasoning in action.