Claude Haiku Glossary
The small, fast tier of Anthropic's Claude family, intended for work where latency and cost matter more than maximum capability.
Claude Haiku is the small, fast tier of Anthropic’s Claude model family. Where Opus prioritises capability and Sonnet balances capability and cost, Haiku is designed for lower latency and lower-priced inference.
Typical uses include classification, extraction, routing, and other high-volume tasks whose answers can be checked cheaply. A harness may also use a smaller model for a first pass before sending harder cases to a stronger tier.
The trade-off is lower capability on difficult reasoning, writing, and coding tasks. Whether Haiku is economical depends on the task, the required reliability, and the cost of retries or escalation.
See also
- Claude — the model family
- Claude Opus, Claude Sonnet — peer tiers
- Sovereign Compute