# Claude 3 Haiku

> Anthropic's fast, low-cost Claude 3 model with vision and a 200K context — now deprecated, retiring April 19, 2026.

- Maker: Anthropic
- Line: Claude Haiku
- Canonical page: https://ai-tldr.dev/models/claude-3-haiku/
- Updated: 2026-08-20

## In short

Claude 3 Haiku is the smallest, fastest, and most cost-effective model in Anthropic's original Claude 3 family, released on March 13, 2024 alongside Claude 3 Sonnet and the flagship Claude 3 Opus. It has a 200K-token context window and costs $0.25 per million input tokens and $1.25 per million output tokens. You can use it through Anthropic API, Amazon Bedrock and Google Cloud Vertex AI.

## Facts

| Fact | Value | Source |
| --- | --- | --- |
| Context window | 200K tokens (max output 4,096 tokens) | [1] |
| Input price | $0.25 / 1M tokens | [2] |
| Output price | $1.25 / 1M tokens | [2] |
| Licence | Proprietary · API only | [1] |
| Released | 13 March 2024 | [3] |
| Parameters | Undisclosed | [1] |
| Knowledge cutoff | August 2023 | [1] |

## Sources

1. [Anthropic — Claude 3 model card (PDF)](https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf)
2. [Anthropic — Claude 3 Haiku announcement](https://www.anthropic.com/news/claude-3-haiku)
3. [Anthropic — Introducing the Claude 3 family](https://www.anthropic.com/news/claude-3-family)

## Overview

Claude 3 Haiku is the smallest, fastest, and most cost-effective model in Anthropic's original Claude 3 family, released on March 13, 2024 alongside Claude 3 Sonnet and the flagship Claude 3 Opus. Served via the API as `claude-3-haiku-20240307`, it was positioned as a near-instant model for high-volume, latency-sensitive work — Anthropic described it as processing roughly 21K tokens (about 30 pages) per second for prompts under 32K tokens.

Like the rest of the Claude 3 line, Claude 3 Haiku is multimodal: it accepts both text and images (photos, charts, graphs, and technical diagrams) and returns text. It ships with a 200K-token context window and a knowledge cutoff of August 2023. Pricing was set at $0.25 per million input tokens and $1.25 per million output tokens — Anthropic illustrated the value by noting Haiku could analyze about 400 Supreme Court cases or 2,500 images for roughly one US dollar.

Claude 3 Haiku is now deprecated. Anthropic has scheduled `claude-3-haiku-20240307` for retirement on April 19, 2026, and recommends migrating to Claude Haiku 4.5 (`claude-haiku-4-5`), which offers far higher intelligence, a larger 200K context with 64K max output, and broader feature support. It sits at the base of the Claude Haiku lineage that runs through Claude 3.5 Haiku to Claude Haiku 4.5.

## Benchmarks

- MMLU (5-shot): 75.2% ([source](https://www.vellum.ai/blog/llm-benchmarks-overview-limits-and-model-comparison))
- HumanEval (code): 75.9% ([source](https://www.vellum.ai/blog/llm-benchmarks-overview-limits-and-model-comparison))
- GPQA (reasoning): 35.7% ([source](https://www.vellum.ai/blog/llm-benchmarks-overview-limits-and-model-comparison))
- MATH: 38.9% ([source](https://www.vellum.ai/blog/llm-benchmarks-overview-limits-and-model-comparison))
- MGSM (multilingual math): 71.7% ([source](https://www.vellum.ai/blog/llm-benchmarks-overview-limits-and-model-comparison))

## Pricing

- Input: $0.25 / 1M tokens
- Output: $1.25 / 1M tokens

Launch pricing for claude-3-haiku-20240307.

Source: https://www.anthropic.com/news/claude-3-haiku

## Strengths

- Very low cost — $0.25 input / $1.25 output per million tokens at launch, the cheapest Claude 3 tier
- Fast, near-instant responses suited to high-volume and latency-sensitive workloads
- Multimodal: accepts text and images (charts, diagrams, photos) with text output
- Large 200K-token context window for a budget-tier model
- Available across the Claude API, Amazon Bedrock, and Google Cloud Vertex AI

## Best for

- High-volume, cost-sensitive tasks like classification, tagging, and content moderation
- Customer-support and chat assistants needing fast, cheap responses
- Bulk summarization and extraction over large documents (within the 200K context)
- Basic image understanding — reading charts, diagrams, and scanned documents
- Lightweight coding help and simple data processing at scale

## How to access

- [Anthropic API](https://docs.anthropic.com/en/docs/about-claude/models) — model id `claude-3-haiku-20240307`
- [Amazon Bedrock](https://aws.amazon.com/bedrock/claude/) — model id `anthropic.claude-3-haiku-20240307-v1:0`
- [Google Cloud Vertex AI](https://cloud.google.com/vertex-ai/generative-ai/docs/partner-models/claude) — model id `claude-3-haiku@20240307`

## Claude Haiku versions

- [Claude Haiku 4.5](https://ai-tldr.dev/models/claude-haiku-4-5/) (15 October 2025)
- [Claude 3.5 Haiku](https://ai-tldr.dev/models/claude-3-5-haiku/) (22 October 2024)
- Claude 3 Haiku (13 March 2024) — this page

## FAQ

### Is Claude 3 Haiku still available?

It is deprecated. Anthropic has scheduled the model ID claude-3-haiku-20240307 for retirement on April 19, 2026, after which API calls to it will fail. Anthropic recommends migrating to Claude Haiku 4.5 (claude-haiku-4-5).

### How much did Claude 3 Haiku cost?

At launch it was priced at $0.25 per million input tokens and $1.25 per million output tokens — a 1:5 input-to-output ratio and the cheapest tier in the Claude 3 family.

### What is Claude 3 Haiku's context window and knowledge cutoff?

It has a 200K-token context window, a maximum output of 4,096 tokens, and a training knowledge cutoff of August 2023.

### Can Claude 3 Haiku process images?

Yes. Like all Claude 3 models, Haiku is multimodal — it accepts text and images (photos, charts, graphs, and technical diagrams) and returns text.

## Links

- [Anthropic — Claude 3 Haiku announcement](https://www.anthropic.com/news/claude-3-haiku)
- [Anthropic — Introducing the Claude 3 family](https://www.anthropic.com/news/claude-3-family)
- [Anthropic — Claude 3 model card (PDF)](https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf)
- [Anthropic — Models overview & deprecations](https://docs.anthropic.com/en/docs/about-claude/models)
- [Anthropic API — getting started](https://docs.anthropic.com/en/api/getting-started)
- [Vellum — LLM benchmark comparison (Claude 3 scores)](https://www.vellum.ai/blog/llm-benchmarks-overview-limits-and-model-comparison)
- [Artificial Analysis — Claude 3 Haiku analysis](https://artificialanalysis.ai/models/claude-3-haiku)
- [Anthropic model deprecations (retirement dates)](https://platform.claude.com/docs/en/about-claude/model-deprecations)

Source: AI/TLDR LLM registry, https://ai-tldr.dev/models/claude-3-haiku/