Claude models explained: Haiku, Sonnet, Opus or Fable
TL;DR Sonnet covers most work. Turn the effort dial down before you pay for a bigger model.
Haiku, Sonnet, Opus or Fable? Most teams pick by vibes and overpay. Here's my cheat sheet.
- 💨 Haiku: fast and cheap. Sorting, tagging, small jobs inside a bigger agent
- ⚡ Sonnet: the everyday default for coding and analysis. For most products, the only model you need
- 🧠 Opus: the hard stuff. Long-running agents, big refactors
- 🚀 Fable: the top. Only when Opus already fell short
Try this
Before you upgrade a model: turn on prompt caching, batch what can wait, then lower the effort dial. That's where the money is.
