Qwen2.5 3B: Complete Benchmark Performance Guide

alibaba 2.5-3b 2024-09-19 Qwen License

In-depth analysis of Qwen2.5 3B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendoralibaba
Version2.5-3b
Release Date2024-09-19
Context Window32768 tokens
Input Modalitiestext
Output Modalitiestext
LicenseQwen License
Documentationhttps://qwenlm.github.io/

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU56.8%2024-09-195-shotview
HUMANEVAL49pass@12024-09-19—view
GSM8K31.3%2024-09-190-shot CoTview
MATH17%2024-09-190-shot CoTview
BBH48.8%2024-09-193-shot CoTview
GPQA18%2024-09-190-shotview
IFEVAL56%2024-09-19prompt_strictview
ARC76.5%2024-09-19challengeview
MUSR23.3%2024-09-190-shotview
WINOGRANDE61.4%2024-09-190-shotview

Pricing

TierPriceCurrency
Input$0.1 / MtokUSD
Output$0.1 / 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 3B

모델 개요

阿里巴巴 Qwen2.5 3B 轻量开源模型, 32K 上下文, 3B 参数, 适合移动设备与边缘部署。

핵심 사양

공급업체버전출시일컨텍스트 창입력 모달리티출력 모달리티라이선스
Alibaba2.5-3b2024-09-1932KtexttextQwen License

벤치마크 성능

벤치마크점수단위비고
MMLU (Massive Multitask Language Understanding)56.8%5-shot
HumanEval49.0pass@1—
GSM8K (Grade School Math 8K)31.3%0-shot CoT
MATH17.0%0-shot CoT
BBH (BIG-Bench Hard)48.8%3-shot CoT
GPQA18.0%0-shot
IFEval56.0%prompt_strict
ARC76.5%challenge
MUSR23.3%0-shot
WinoGrande61.4%0-shot

가격

입력출력캐시 읽기캐시 쓰기
————

백만 토큰당

강점

  • 可靠的通用模型。

약점

  • MMLU 仅 56.8，知识推理偏弱。
  • HumanEval 49.0，代码能力较弱。
  • 闭源专有模型，不支持自托管。

사용 사례

  • 通用对话与问答

참고문헌