Claude 3 Haiku: Complete Benchmark Performance Guide

anthropic 3-haiku 2024-03-07 Proprietary

In-depth analysis of Claude 3 Haiku performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendoranthropic
Version3-haiku
Release Date2024-03-07
Context Window200000 tokens
Input Modalitiestext, image
Output Modalitiestext
LicenseProprietary
Documentationhttps://docs.anthropic.com/claude/docs

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 77.9 % 2024-03-07 5-shot view
HUMANEVAL 69.4 pass@1 2024-03-07 — view
GSM8K 82.2 % 2024-03-07 0-shot CoT view
MATH 52.2 % 2024-03-07 0-shot CoT view
BBH 75.7 % 2024-03-07 3-shot CoT view
GPQA 44.4 % 2024-03-07 0-shot view
IFEVAL 70.8 % 2024-03-07 prompt_strict view
ARC 94.3 % 2024-03-07 challenge view
MUSR 58.1 % 2024-03-07 0-shot view
WINOGRANDE 84.8 % 2024-03-07 0-shot view

Pricing

Tier Price Currency
Input$0.25 / MtokUSD
Output$1.25 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://www.anthropic.com/pricing · as of 2024-03-07

Compliance

  • Data Residency: US
  • SOC2: ✓
  • HIPAA: ✗
  • GDPR: ✓
  • ISO 27001: ✗

Claude 3 Haiku

Model Overview

Anthropic Claude 3 Haiku 经济型模型, 200K 上下文, 响应速度最快, 适合高吞吐量实时场景。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Anthropic 3-haiku 2024-03-07 200K text, image text Proprietary

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 77.9 % 5-shot
HumanEval 69.4 pass@1 —
GSM8K (Grade School Math 8K) 82.2 % 0-shot CoT
MATH 52.2 % 0-shot CoT
BBH (BIG-Bench Hard) 75.7 % 3-shot CoT
GPQA 44.4 % 0-shot
IFEval 70.8 % prompt_strict
ARC 94.3 % challenge
MUSR 58.1 % 0-shot
WinoGrande 84.8 % 0-shot

Pricing

Input Output Cache Read Cache Write
— — — —

per million tokens

Strengths

  • Reliable general-purpose model.

Weaknesses

  • Proprietary, not self-hostable.

Use Cases

  • General chat and Q&A

References