Llama 2-70B
Code source ouvertpar Meta · Sorti le 2024-01-01
56.6
score moyen
N/A
Prix d'entrée
N/A
Prix de sortie
N/A
Fenêtre de contexte
text
Type
Tested on 12 benchmarks with 56.6% average. Top scores: TriviaQA (87.6%), HellaSwag (80.4%), LAMBADA (78.9%).
Scores de benchmark
| Benchmark | Catégorie | Score | Bar |
|---|---|---|---|
| TriviaQA | knowledge | 87.6 | |
| HellaSwag | knowledge | 80.4 | |
| LAMBADA | knowledge | 78.9 | |
| ARC AI2 | knowledge | 71.1 | |
| GSM8K | math | 69.6 | |
| PIQA | knowledge | 65.6 | |
| Winogrande | knowledge | 60.4 | |
| MMLU | knowledge | 59.9 | |
| BBH | reasoning | 53.2 | |
| OpenBookQA | knowledge | 46.9 | |
| MATH level 5 | math | 3.3 | |
| GPQA diamond | knowledge | 1.8 |