Llama 3.3 70B Instruct vs Qwen2.5 32B Instruct
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | Llama 3.3 70B Instruct | Qwen2.5 32B Instruct |
|---|---|---|
| Provider |  Meta |  Alibaba |
| puntuación promedio | 46.9 | 43.2 |
| Precio de entrada | $0.10 | - |
| Precio de salida | $0.32 | - |
| Ventana de contexto | 131K tokens (~66 books) | - |
| Publicado el | 2024-12-06 | 2024-09-17 |
| Código abierto | Open Source | Open Source |
Puntuaciones de benchmark
6 benchmarks · Llama 3.3 70B Instruct: 3, Qwen2.5 32B Instruct: 3
| Benchmark | Categoría | Llama 3.3 70B Instruct | Qwen2.5 32B Instruct |
|---|---|---|---|
| BBH (HuggingFace) | general | 56.6 | 56.5 |
| GPQA | knowledge | 10.5 | 11.7 |
| IFEval | language | 90.0 | 83.5 |
| MATH Level 5 | math | 48.3 | 62.5 |
| MMLU-PRO | knowledge | 48.1 | 51.9 |
| MUSR | reasoning | 15.6 | 13.5 |