Model tiers (Haiku / Sonnet / Opus)

Sizes of the same model, sold small to large. Anthropic — the company behind Claude, which you may well already use — names its three Haiku, Sonnet and Opus: fast and cheap, balanced middle, and most capable but priciest. Every other AI company sells the same ladder under its own names.

Small, medium, large. Nobody buys the large one for a job the small one does perfectly well.

Why it matters: the standard money-saving rule is "cheapest one first" — start every task on the small tier and only move up when it visibly fails. Ten runs on the small tier can cost less than one run on the large one, which is the difference between an AI bill you can ignore and one you can't.

Related: Model · Token cost / cost discipline