Qwen2 72B: Complete Benchmark Performance Guide

alibaba 2-72b 2024-06-07 Qwen License

In-depth analysis of Qwen2 72B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendoralibaba
Version2-72b
Release Date2024-06-07
Context Window131072 tokens
Input Modalitiestext
Output Modalitiestext
LicenseQwen License
Documentationhttps://qwenlm.github.io/

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU80.2%2024-06-075-shotview
HUMANEVAL76.5pass@12024-06-07—view
GSM8K87%2024-06-070-shot CoTview
MATH71.9%2024-06-070-shot CoTview
BBH83.4%2024-06-073-shot CoTview
GPQA53.6%2024-06-070-shotview
IFEVAL76.9%2024-06-07prompt_strictview
ARC94.4%2024-06-07challengeview
MUSR68.9%2024-06-070-shotview
WINOGRANDE82.5%2024-06-070-shotview

Pricing

TierPriceCurrency
Input$0.5 / MtokUSD
Output$0.8 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://help.aliyun.com/zh/model-studio/getting-started/models · as of 2024-06-07

Compliance

  • Data Residency: CN
  • SOC2: ✗
  • HIPAA: ✗
  • GDPR: ✗
  • ISO 27001: ✗

Qwen2 72B

Tổng quan mô hình

阿里巴巴 Qwen2 72B 开源旗舰, 131K 上下文, 多语言/数学/编程能力突出, 性能接近 GPT-4。

Thông số kỹ thuật cốt lõi

Nhà cung cấpPhiên bảnNgày phát hànhCửa sổ ngữ cảnhPhương thức đầu vàoPhương thức đầu raGiấy phép
Alibaba2-72b2024-06-07131KtexttextQwen License

Hiệu suất benchmark

BenchmarkĐiểmĐơn vịGhi chú
MMLU (Massive Multitask Language Understanding)80.2%5-shot
HumanEval76.5pass@1—
GSM8K (Grade School Math 8K)87.0%0-shot CoT
MATH71.9%0-shot CoT
BBH (BIG-Bench Hard)83.4%3-shot CoT
GPQA53.6%0-shot
IFEval76.9%prompt_strict
ARC94.4%challenge
MUSR68.9%0-shot
WinoGrande82.5%0-shot

Giá

Đầu vàoĐầu raĐọc bộ nhớ đệmGhi bộ nhớ đệm
————

mỗi triệu token

Điểm mạnh

  • MMLU score 80.2, strong knowledge reasoning.
  • GSM8K 87.0, robust math reasoning.

Điểm yếu

  • 闭源专有模型，不支持自托管。

Trường hợp sử dụng

  • 代码生成与调试

Tham chiếu