Key Specifications
| Vendor | alibaba |
|---|
| Version | 2-72b |
|---|
| Release Date | 2024-06-07 |
|---|
| Context Window | 131072 tokens |
|---|
| Input Modalities | text |
|---|
| Output Modalities | text |
|---|
| License | Qwen License |
|---|
| Documentation | https://qwenlm.github.io/ |
|---|
Benchmark Performance
| Benchmark | Score | Unit | Evaluated At | Notes | Source |
|---|
| MMLU | 80.2 | % | 2024-06-07 | 5-shot | view |
| HUMANEVAL | 76.5 | pass@1 | 2024-06-07 | — | view |
| GSM8K | 87 | % | 2024-06-07 | 0-shot CoT | view |
| MATH | 71.9 | % | 2024-06-07 | 0-shot CoT | view |
| BBH | 83.4 | % | 2024-06-07 | 3-shot CoT | view |
| GPQA | 53.6 | % | 2024-06-07 | 0-shot | view |
| IFEVAL | 76.9 | % | 2024-06-07 | prompt_strict | view |
| ARC | 94.4 | % | 2024-06-07 | challenge | view |
| MUSR | 68.9 | % | 2024-06-07 | 0-shot | view |
| WINOGRANDE | 82.5 | % | 2024-06-07 | 0-shot | view |
Compliance
- Data Residency: CN
- SOC2: ✗
- HIPAA: ✗
- GDPR: ✗
- ISO 27001: ✗
Qwen2 72B
Modellübersicht
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。
Kernspezifikationen
| Anbieter | Version | Veröffentlichungsdatum | Kontextfenster | Eingabemodalitäten | Ausgabemodalitäten | Lizenz |
|---|
| Alibaba | 2-72b | 2024-06-07 | 131K | text | text | Qwen License |
Benchmark-Leistung
| Benchmark | Ergebnis | Einheit | Notizen |
|---|
| MMLU (Massive Multitask Language Understanding) | 80.2 | % | 5-shot |
| HumanEval | 76.5 | pass@1 | — |
| GSM8K (Grade School Math 8K) | 87.0 | % | 0-shot CoT |
| MATH | 71.9 | % | 0-shot CoT |
| BBH (BIG-Bench Hard) | 83.4 | % | 3-shot CoT |
| GPQA | 53.6 | % | 0-shot |
| IFEval | 76.9 | % | prompt_strict |
| ARC | 94.4 | % | challenge |
| MUSR | 68.9 | % | 0-shot |
| WinoGrande | 82.5 | % | 0-shot |
Preise
| Eingabe | Ausgabe | Cache-Lesen | Cache-Schreiben |
|---|
| — | — | — | — |
pro Million Token
Stärken
- MMLU score 80.2, strong knowledge reasoning.
- GSM8K 87.0, robust math reasoning.
Schwächen
Anwendungsfälle
Referenzen