Qwen2-72B
Open Sourcedi Alibaba Qwen · Rilascio 2024-01-01
41.3
punteggio medio
N/A
Prezzo Input
N/A
Prezzo Output
N/A
Finestra di Contesto
text
Tipo
Tested on 12 benchmarks with 41.3% average. Top scores: CMMLU (89.7%), MMLU (76.5%), Aider — Code Editing (55.6%).
Punteggi Benchmark
| Benchmark | Categoria | Punteggio | Bar |
|---|---|---|---|
| CMMLU | knowledge | 89.7 | |
| MMLU | knowledge | 76.5 | |
| Aider — Code Editing | coding | 55.6 | |
| MMLU-PRO | knowledge | 52.6 | |
| BBH (HuggingFace) | general | 51.9 | |
| MATH level 5 | math | 39.1 | |
| IFEval | language | 38.2 | |
| MATH Level 5 | math | 31.1 | |
| GPQA diamond | knowledge | 21.0 | |
| MUSR | reasoning | 19.7 | |
| GPQA | knowledge | 19.2 | |
| The Agent Company | agentic | 1.1 |