Claude 3 Sonnet: Complete Benchmark Performance Guide

anthropic 3-sonnet 2024-03-04 Proprietary

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

Key Specifications

Vendoranthropic
Version3-sonnet
Release Date2024-03-04
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 81.5 % 2024-03-04 5-shot view
HUMANEVAL 74.5 pass@1 2024-03-04 — view
GSM8K 84 % 2024-03-04 0-shot CoT view
MATH 42.4 % 2024-03-04 0-shot CoT view
BBH 76.1 % 2024-03-04 3-shot CoT view
GPQA 35.9 % 2024-03-04 0-shot view
IFEVAL 74.3 % 2024-03-04 prompt_strict view
ARC 94.8 % 2024-03-04 challenge view
MUSR 64.1 % 2024-03-04 0-shot view
WINOGRANDE 83.8 % 2024-03-04 0-shot view

Pricing

Tier Price Currency
Input$3 / MtokUSD
Output$15 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://www.anthropic.com/pricing · as of 2024-03-04

Compliance

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

Claude 3 Sonnet

Model Overview

Anthropic Claude 3 Sonnet 平衡型模型, 200K 上下文, 在速度与智能间取得平衡, 适合企业级部署。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Anthropic 3-sonnet 2024-03-04 200K text, image text Proprietary

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 81.5 % 5-shot
HumanEval 74.5 pass@1 —
GSM8K (Grade School Math 8K) 84.0 % 0-shot CoT
MATH 42.4 % 0-shot CoT
BBH (BIG-Bench Hard) 76.1 % 3-shot CoT
GPQA 35.9 % 0-shot
IFEval 74.3 % prompt_strict
ARC 94.8 % challenge
MUSR 64.1 % 0-shot
WinoGrande 83.8 % 0-shot

Pricing

Input Output Cache Read Cache Write
— — — —

per million tokens

Strengths

  • MMLU score 81.5, strong knowledge reasoning.
  • Input Modalities: text, image, audio.
  • Context window 200K.

Weaknesses

  • Proprietary, not self-hostable.

Use Cases

  • Code generation and debugging
  • Long document summarization
  • Vision and image understanding

References