Llama 3 8B Instruct vs Qwen2-72B
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | Llama 3 8B Instruct | Qwen2-72B |
|---|---|---|
| Provider |  Meta |  Alibaba Qwen |
| puntuación promedio | 30.8 | 41.3 |
| Precio de entrada | $0.03 | - |
| Precio de salida | $0.04 | - |
| Ventana de contexto | 8K tokens (~4 books) | - |
| Publicado el | 2024-04-18 | 2024-01-01 |
| Código abierto | Open Source | Open Source |
Puntuaciones de benchmark
9 benchmarks · Llama 3 8B Instruct: 1, Qwen2-72B: 8
| Benchmark | Categoría | Llama 3 8B Instruct | Qwen2-72B |
|---|---|---|---|
| BBH (HuggingFace) | general | 18.4 | 51.9 |
| GPQA | knowledge | 2.1 | 19.2 |
| GPQA diamond | knowledge | 1.4 | 21.0 |
| IFEval | language | 24.0 | 38.2 |
| MATH level 5 | math | 6.1 | 39.1 |
| MATH Level 5 | math | 3.9 | 31.1 |
| MMLU | knowledge | 58.4 | 76.5 |
| MMLU-PRO | knowledge | 17.8 | 52.6 |
| MUSR | reasoning | 19.9 | 19.7 |