CONFIRMEDDate confidence: CONFIRMEDStated as firmHow firmly the source words the date: confirmed, scheduled, estimated, delayed or unverified.Show all CONFIRMED pins90% CONFIDENCEOverall confidence: 90%How well the pin's source and references back up its dates.Weighted average of how firmly 5 references, the source included, support the pin's start and end times; a reference counts half as much for every 180 days older than the newestShow all pins at 75% confidence or better
Anthropic's[1][5] post is dated "Mar 13, 2024": "Today we're releasing Claude[2] 3 Haiku ... The model is now available alongside Sonnet and Opus in the Claude API and on claude.ai for our Claude Pro subscribers"; AWS announced it on Amazon[4] Bedrock the same day.- Start
- Mar 13, 202490% CONFIDENCEsource
Claude 3 Haiku Released
Anthropic releases Claude 3 Haiku, the fastest and cheapest Claude 3 model at $0.25 per million input tokens, able to read about 30 pages of text per second, nine days after announcing it with Opus and Sonnet.
- Anthropic announced Claude 3 Haiku with Opus and Sonnet on 4 March 2024 as "available soon", and released it on 13 March in the Claude API and on claude.ai for Claude Pro subscribers[1][5]
- It launched on Amazon Bedrock the same day, with Google Cloud Vertex AI "coming soon"[1][4]
- Anthropic pitches it at enterprises that need to analyze large volumes of quarterly filings, contracts or legal cases "for half the cost of other models in its performance tier"[1][3]
- The model stayed on the Claude API for just over two years: deprecated on 19 February 2026 and retired on 20 April 2026 in favour of Claude Haiku 4.5[2]
Notable features
- The fastest and most compact Claude 3 model, priced at $0.25 per million input tokens and $1.25 per million output tokens - a 1:5 ratio built for long-prompt enterprise work[1][5]
- Processes 21K tokens (about 30 pages) per second for prompts under 32K tokens, "three times faster than its peers for the vast majority of workloads"; one US dollar buys analysis of 400 Supreme Court cases or 2,500 images[1][3]
- Benchmarks Anthropic reports: 75.2% on MMLU, 75.9% on HumanEval, 88.9% on GSM8K and 33.3% on GPQA Diamond, ahead of GPT-3.5 and Gemini 1.0 Pro on most rows[1]
- A 200K-token context window and the Claude 3 family's vision, reading photos, charts and diagrams[1][4]
Benchmarks[1]
| Benchmark | Claude 3 Haiku | GPT-3.5 | Gemini 1.0 Pro |
|---|---|---|---|
| Price, input / output | $0.25 / $1.25 | $0.50 / $1.50 | $0.47 / $1.42 |
| MMLU (undergraduate level knowledge) | 75.2% 5-shot | 70.0% 5-shot | 71.8% 5-shot |
| GPQA, Diamond (graduate level reasoning) | 33.3% 0-shot CoT | 28.1% 0-shot CoT | — |
| GSM8K (grade school math) | 88.9% 0-shot CoT | 57.1% 5-shot | 86.5% Maj1@32 |
| MATH (math problem-solving) | 40.9% 4-shot | 34.1% 4-shot | 32.6% 4-shot |
| MGSM (multilingual math) | 76.5% 8-shot | — | 63.5% 8-shot |
| HumanEval (code) | 75.9% 0-shot | 48.1% 0-shot | 67.7% 0-shot |
| DROP, F1 score (reasoning over text) | 78.4 3-shot | 64.1 3-shot | 74.1 Variable shots |
| BIG-Bench-Hard (mixed evaluations) | 73.7% 3-shot CoT | 66.6% 3-shot CoT | 75.0% 3-shot CoT |
| ARC-Challenge (knowledge Q&A) | 89.2% 25-shot | 85.2% 25-shot | — |
| HellaSwag (common knowledge) | 85.9% 10-shot | 85.5% 10-shot | 84.7% 10-shot |
| Vision-enabled | Yes | No | Yes |
Buy onClaude