About
Claude 3 Haiku is Anthropic's fastest and most compact model for near-instant responsiveness. Quick and accurate targeted performance. See the launch announcement and benchmark results [here](https://www.anthropic.com/news/claude-3-haiku) #multimodal
Related Models
Anthropic: Claude Fable 52026-06-09Anthropic: Claude Opus 4.82026-05-27Anthropic: Claude Opus 4.8 (Fast)2026-05-27Anthropic: Claude Opus 4.7 (Fast)2026-05-12Anthropic: Claude Opus 4.72026-04-16Claude Opus 4.7 (Non-reasoning, High Effort)2026-04-16Anthropic: Claude Opus 4.6 (Fast)2026-04-07Claude Sonnet 4.6 (Adaptive Reasoning, Max Effort)2026-02-17
Pricing
Input
$0.25
per 1M tokens
Output
$1.25
per 1M tokens
Blended
$0.50
per 1M tokens
Cheaper than 52% of models. Median price is $0.54/1M tokens.
Cost Calculator
Tokens per day1M
100K100M
Daily
$0.50
Monthly
$15.00
vs. Similar Models
Olmo 3 7B ThinkQ:+0.1
$0.00-100%
Reka Flash 3Q:+0.2
$0.13-75%
GPT-3.5 TurboQ:-0.3
$0.75+50%
Mistral MediumQ:-0.3
$4.09+718%
Performance
134
tokens/sec
Faster than 66% of models
0.52
seconds
Faster than 86% of models
0.52
seconds
Faster than 91% of models
Market Median
94 tok/s
42% faster
Median TTFT
1.11s
54% faster
Throughput/Dollar
269
tok/s per $/1M
Speed Comparison
OpenAI: GPT-4.1
134 tok/s-0%
Grok 4.3 (low)
134 tok/s-0%
Qwen3 VL 8B (Reasoning)
135 tok/s+1%
Context Window
200K
tokens
Larger than 50% of models
Max Output
4K
tokens
2% of context
Benchmarks
MMLU-ProNot evaluated
GPQA Diamond
37.4%
HLE
3.9%
LiveCodeBench
15.4%
SciCode
18.6%
TerminalBench Hard
0.8%
MATH-500
39.4%
AIME
1.0%
AIME 2025Not evaluated
IFBench
36.1%
Long Context Recall
21.0%
Tau2
21.1%
Market AverageTop Score