Brain labTry it → read → next · ~8 min

Tutorials · Chapter C (3/4) · ~8 min

Reasoning models

Try it → see it → read → next

Some models spend extra compute **thinking stepwise** — great for hard logic, overkill for simple rewrites.

On paths: Student

Optional on: Professional

Try yourself

Playground

Reasoning vs fast model

Reasoning models cost more latency — use them when stepwise logic beats speed.

Multi-step budget word problem

Recap

What you just did

You sorted tasks into reasoning vs fast buckets — same eval mindset as benchmarks.

Read

When to reach for reasoning

  • Multi-step math, planning, constraint puzzles
  • When cheap models fail your golden eval set

Read

When to stay fast

  • Classification, formatting, short drafts
  • High-volume paths where [p95 latency](/reference/p95-latency) matters