Qwen1.5 32B: Complete Benchmark Performance Guide

alibaba 1.5-32b 2024-02-25 Qwen License

In-depth analysis of Qwen1.5 32B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendoralibaba
Version1.5-32b
Release Date2024-02-25
Context Window32768 tokens
Input Modalitiestext
Output Modalitiestext
LicenseQwen License
Documentationhttps://qwenlm.github.io/

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 81.9 % 2024-02-25 5-shot view
HUMANEVAL 77.9 pass@1 2024-02-25 — view
GSM8K 81.1 % 2024-02-25 0-shot CoT view
MATH 46.3 % 2024-02-25 0-shot CoT view
BBH 73 % 2024-02-25 3-shot CoT view
GPQA 30.6 % 2024-02-25 0-shot view
IFEVAL 68.2 % 2024-02-25 prompt_strict view
ARC 91.4 % 2024-02-25 challenge view
MUSR 54.5 % 2024-02-25 0-shot view
WINOGRANDE 79.6 % 2024-02-25 0-shot view

Pricing

Tier Price Currency
Input$0.35 / MtokUSD
Output$0.45 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://help.aliyun.com/zh/model-studio/getting-started/models · as of 2024-02-25

Compliance

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

Qwen1.5 32B

Model Overview

阿里巴巴 Qwen1.5 32B 中端开源模型, 32K 上下文, 性能接近 72B, 适合中等规模部署。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Alibaba 1.5-32b 2024-02-25 32K text text Qwen License

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 81.9 % 5-shot
HumanEval 77.9 pass@1 —
GSM8K (Grade School Math 8K) 81.1 % 0-shot CoT
MATH 46.3 % 0-shot CoT
BBH (BIG-Bench Hard) 73.0 % 3-shot CoT
GPQA 30.6 % 0-shot
IFEval 68.2 % prompt_strict
ARC 91.4 % challenge
MUSR 54.5 % 0-shot
WinoGrande 79.6 % 0-shot

Pricing

Input Output Cache Read Cache Write
— — — —

per million tokens

Strengths

  • MMLU score 81.9, strong knowledge reasoning.

Weaknesses

  • Proprietary, not self-hostable.

Use Cases

  • Code generation and debugging

References