Claude Haiku: Claude's fast, affordable tier

What Haiku is built for

Haiku shines where latency and cost matter more than depth: classification, information extraction, tagging, moderation, short replies, a first pass over large volumes. In a pipeline it often does the rough work or routing before a higher tier handles the hard cases.

The deliberate trade-off

In exchange for speed and economy, Haiku is less comfortable than Sonnet or Opus on complex reasoning, hard code or very nuanced analysis. The simple rule: if the task is hard or loosely specified, move up a tier; if it is simple, repetitive or massive, Haiku is often the best value.

Haiku at scale

It is the go-to tier for real-time, high-throughput use, where every millisecond and every cent count at scale: assistants embedded in a product, batch processing, background features. A common architecture combines Haiku for the bulk and a more capable model for the exceptions.

Tracking Haiku versions

Like Opus and Sonnet, 'Haiku' refers to a tier, not a fixed version: Anthropic ships new generations that improve its capabilities. Speed, context window and availability change. For the real state and pricing, check Anthropic's official page and follow our Models category.

What is Claude Haiku?

Haiku is the fastest, most affordable tier of Anthropic's Claude family. It is built for volume, real-time, classification and simple tasks at scale, where latency and cost matter more than the deepest reasoning.

Is Haiku the cheapest Claude model?

Yes, Haiku is the most affordable and fastest tier. Exact amounts change; check Anthropic's official pricing.

When should I prefer Haiku over Sonnet?

For large volumes, real time or simple tasks. For reasoning, hard code or deep analysis, move up to Sonnet or Opus.

Can Haiku write code?

It handles simple, fast code, but for hard or multi-file tasks, Sonnet or Opus are better suited.