Claude Haiku 3

Retired Apr 2026

Anthropic's smallest Claude 3 model, released March 2024 and designed for high-throughput tasks like classification and data extraction.

Context: 200K tokens
Max output: 4K
Cutoff: Aug 2023
Released: Mar 2024

Visit Anthropic Compare vs other LLMs →

Packet·ai

B200 @ $3.75/h on demand

No contract, no spot interrupts, no cluster minimums

Full root SSH, web dashboard, CI/CD ready

Reserve Now

Capabilities

What Claude Haiku 3 accepts and can do, as published by Anthropic.

Text Accepts and generates natural-language text.
Image Accepts images as input alongside text.
Function calling Connect to external tools, APIs, and systems.
Structured output Return responses in structured formats like JSON.

Common questions

What is Claude Haiku 3 good for?

High-volume classification and data extraction. It took text and images, and was the cheapest of the Claude 3 generation.

When is Claude Haiku 3 not a good fit?

Any new work: Anthropic retired it in April 2026. Closed weights, so no self-hosting, and its short maximum output and old knowledge kept it to brief answers about older material.

How does Claude Haiku 3 compare with Claude Haiku 3.5?

Anthropic says Claude Haiku 3.5 keeps the speed of Claude Haiku 3 but improves across every skill, with coding as its strong point, better instruction following and more accurate tool use. Same 200K-token context and text and image input. Claude Haiku 3 came out 7 months earlier and has an older knowledge cutoff (August 2023 against July 2024) and a smaller max output (4K against 8K tokens).

Can I self-host Claude Haiku 3?

No. Anthropic does not publish the weights; Claude Haiku 3 is available only through hosted APIs.