Sapiens
Technicals

What is AI reasoning?

Published June 1, 2026 · 4 min read

AI REASONING It shows its work, then circles the answer. A reasoning model fills the margin before committing to the result. STANDARD MODEL · FAST 42 blank margin — just the answer REASONING MODEL · SLOWER, ACCURATE step-by-step working 42 Both reach the same answer — the reasoning model gets there by working through the steps first.

Definition

AI reasoning is when an AI model works through a problem step by step before giving an answer, rather than producing a response in a single instant pass.

At a glance

  • A standard model answers in one quick pass; a reasoning model first works through hidden steps, then answers[5].
  • The hidden working is called chain-of-thought; the extra effort per question is test-time compute[1].
  • It helps most on multi-step tasks (math, planning, analysis) and little on simple lookups[3].
  • The cost is real: answers can run 20 to 80 percent slower and pricier per query.

How it works

A standard model blurts out a plausible answer in one fast pass. A reasoning model pauses to break the problem into steps, weigh options, and check itself first[2]. More thinking generally means better answers on hard problems.

When to use it

Use reasoning for genuinely complex work; keep a fast standard model for quick facts and short replies. Models vary hugely in price and speed, so a common pattern is routing: cheap model for the easy majority, reasoning model only for the few hard questions[4].

Bottom line

Reasoning buys accuracy on hard problems with extra time and money; use it only where the task earns it.

References

  1. What is chain of thought (CoT) prompting? IBM www.ibm.com
  2. The State of LLM Reasoning Model Inference. Sebastian Raschka magazine.sebastianraschka.com
  3. A Visual Guide to Reasoning LLMs. Maarten Grootendorst newsletter.maartengrootendorst.com
  4. Analysis: OpenAI o1 vs DeepSeek R1. Vellum www.vellum.ai
  5. The Ultimate Guide to Reasoning Models. HyScaler hyscaler.com

Comments

Questions, corrections, and links welcome. Be specific and civil.

  • Loading comments…