Qwen2.5 72B Instruct Abliterated vs Llama 3.3 70B Instruct
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | Qwen2.5 72B Instruct Abliterated | Llama 3.3 70B Instruct |
|---|---|---|
| Provider | HA HuiHui AI |  Meta |
| pontuação média | 48.1 | 46.9 |
| Preço de entrada | - | $0.10 |
| Preço de saída | - | $0.32 |
| Janela de contexto | - | 131K tokens (~66 books) |
| Lançado em | 2024-10-26 | 2024-12-06 |
| Código aberto | Proprietary | Open Source |
Pontuações de benchmark
6 benchmarks · Qwen2.5 72B Instruct Abliterated: 4, Llama 3.3 70B Instruct: 2
| Benchmark | Categoria | Qwen2.5 72B Instruct Abliterated | Llama 3.3 70B Instruct |
|---|---|---|---|
| BBH (HuggingFace) | general | 60.5 | 56.6 |
| GPQA | knowledge | 19.4 | 10.5 |
| IFEval | language | 85.9 | 90.0 |
| MATH Level 5 | math | 60.1 | 48.3 |
| MMLU-PRO | knowledge | 50.4 | 48.1 |
| MUSR | reasoning | 12.3 | 15.6 |