Meta Llama 3 8B vs Qwen2.5 72B Instruct
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | Meta Llama 3 8B | Qwen2.5 72B Instruct |
|---|---|---|
| Provider |  Meta |  Alibaba Qwen |
| puntuación promedio | 44.2 | 53.2 |
| Precio de entrada | - | $0.12 |
| Precio de salida | - | $0.39 |
| Ventana de contexto | - | 33K tokens (~16 books) |
| Publicado el | 2024-04-17 | 2024-09-19 |
| Código abierto | Open Source | Open Source |
Puntuaciones de benchmark
6 benchmarks · Meta Llama 3 8B: 2, Qwen2.5 72B Instruct: 4
| Benchmark | Categoría | Meta Llama 3 8B | Qwen2.5 72B Instruct |
|---|---|---|---|
| BBH (HuggingFace) | general | 48.7 | 61.9 |
| GPQA | knowledge | 19.7 | 16.7 |
| IFEval | language | 16.0 | 86.4 |
| MATH Level 5 | math | 18.6 | 59.8 |
| MMLU-PRO | knowledge | 41.2 | 51.4 |
| MUSR | reasoning | 16.0 | 11.7 |