LLaMA-13B vs Llama 3.1 70B Instruct
Côte à côte. Chaque métrique. Chaque benchmark.
| Type | LLaMA-13B | Llama 3.1 70B Instruct |
|---|---|---|
| Provider |  Meta | Meta |
| score moyen | 34.9 | 37.8 |
| Prix d'entrée | - | $0.40 |
| Prix de sortie | - | $0.40 |
| Fenêtre de contexte | - | 131K tokens (~66 books) |
| Sorti le | 2024-01-01 | 2024-07-23 |
| Code source ouvert | Open Source | Open Source |
Scores de benchmark
9 benchmarks · LLaMA-13B: 0, Llama 3.1 70B Instruct: 9
| Benchmark | Catégorie | LLaMA-13B | Llama 3.1 70B Instruct |
|---|---|---|---|
| BBH (HuggingFace) | general | 25.3 | 55.9 |
| Chatbot Arena Elo — Overall | arena | 970.9 | 1292.8 |
| CMMLU | knowledge | 39.8 | 64.4 |
| GPQA | knowledge | 3.5 | 14.2 |
| IFEval | language | 25.3 | 86.7 |
| MATH Level 5 | math | 3.1 | 38.1 |
| MMLU | knowledge | 30.3 | 73.5 |
| MMLU-PRO | knowledge | 23.1 | 47.9 |
| MUSR | reasoning | 2.0 | 17.7 |