Claude 3 Haiku

// 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

1. MMLU (5-shot)75.2%
2. HumanEval (code)75.9%
3. GPQA (reasoning)35.7%
4. MATH38.9%
5. MGSM (multilingual math)71.7%

Scores on a 0–100 scale (25-point gridlines); higher is better. Each benchmark links to its published source.

// Pricing

Launch pricing for claude-3-haiku-20240307.

Pricing source ↗

// 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

// Claude Haiku — every version

The full lineage of the Claude Haiku line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.