DeepSeek R1 Distill Llama 8B vs DeepSeek R1 Distill Qwen 7B
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | DeepSeek R1 Distill Llama 8B | DeepSeek R1 Distill Qwen 7B |
|---|---|---|
| Provider |  DeepSeek | DeepSeek |
| puntuación promedio | 33.6 | 32.7 |
| Precio de entrada | - | - |
| Precio de salida | - | - |
| Ventana de contexto | - | - |
| Publicado el | 2025-01-20 | 2025-01-20 |
| Código abierto | Open Source | Open Source |
Puntuaciones de benchmark
11 benchmarks · DeepSeek R1 Distill Llama 8B: 5, DeepSeek R1 Distill Qwen 7B: 6
| Benchmark | Categoría | DeepSeek R1 Distill Llama 8B | DeepSeek R1 Distill Qwen 7B |
|---|---|---|---|
| BBH (HuggingFace) | general | 5.3 | 7.9 |
| GPQA | knowledge | 0.7 | 3.9 |
| IFEval | language | 37.8 | 40.4 |
| JCommonsenseQA | language | 62.4 | 59.8 |
| JMMLU | language | 37.8 | 42.3 |
| JNLI | language | 69.4 | 54.6 |
| JSQuAD | language | 80.2 | 74.2 |
| LLM-JP — Overall | language | 41.4 | 39.3 |
| MATH Level 5 | math | 22.0 | 19.6 |
| MMLU-PRO | knowledge | 12.1 | 14.7 |
| MUSR | reasoning | 0.5 | 3.5 |