DeepSeek-V2 (MoE-236B, May 2024)
Open Sourcevon DeepSeek · Veroeffentlicht 2024-01-01
76.5
Durchschn. Score
N/A
Eingabepreis
N/A
Ausgabepreis
N/A
Kontextfenster
text
Typ
Tested on 7 benchmarks with 76.5% average. Top scores: ARC AI2 (89.6%), HellaSwag (82.8%), TriviaQA (80.0%).
Benchmark-Ergebnisse
| Benchmark | Kategorie | Score | Bar |
|---|---|---|---|
| ARC AI2 | knowledge | 89.6 | |
| HellaSwag | knowledge | 82.8 | |
| TriviaQA | knowledge | 80.0 | |
| Winogrande | knowledge | 72.6 | |
| BBH | reasoning | 71.7 | |
| MMLU | knowledge | 71.2 | |
| PIQA | knowledge | 67.8 |