Qwen2.5 72B: Complete Benchmark Performance Guide

alibaba 2.5-72b 2024-09-19 Qwen License

In-depth analysis of Qwen2.5 72B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendoralibaba
Version2.5-72b
Release Date2024-09-19
Context Window131072 tokens
Input Modalitiestext
Output Modalitiestext
LicenseQwen License
Documentationhttps://qwenlm.github.io/

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU86.1%2024-09-195-shotview
HUMANEVAL86.6pass@12024-09-19—view
GSM8K88.4%2024-09-190-shot CoTview
MATH83.1%2024-09-190-shot CoTview
BBH82.4%2024-09-193-shot CoTview

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 2024-09-19

Compliance

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

Qwen2.5 72B

Modeloverzicht

阿里巴巴通义千问 Qwen2.5 72B 开源模型，131K 上下文窗口，在中文理解、代码生成与数学推理上表现突出，是当前最强的中文开源模型之一。

Kernspecificaties

LeverancierVersieReleasedatumContextvensterInvoermodaliteitenUitvoermodaliteitenLicentie
Alibaba2.5-72b2024-09-19131KtexttextQwen License

Benchmarkprestaties

BenchmarkScoreEenheidNotities
MMLU (Massive Multitask Language Understanding)86.1%5-shot
HumanEval86.6pass@1—
GSM8K (Grade School Math 8K)88.4%0-shot CoT
MATH83.1%0-shot CoT
BBH (BIG-Bench Hard)82.4%3-shot CoT

Prijzen

InvoerUitvoerCache-lezenCache-schrijven
————

per miljoen tokens

Sterktes

  • MMLU score 86.1, strong knowledge reasoning.
  • HumanEval 86.6, excellent code generation.
  • GSM8K 88.4, robust math reasoning.

Zwaktes

  • 闭源专有模型，不支持自托管。

Gebruiksscenario’s

  • 代码生成与调试

Referenties