DeepSeek V3
C贸digo abiertopor DeepSeek 路 Publicado el 2024-12-26
57.2
puntuaci贸n promedio
$0.32/1M
Precio de entrada
$0.89/1M
Precio de salida
164K tokens (~82 books)
Ventana de contexto
text
Tipo
Tested on 13 benchmarks with 57.2% average. Top scores: ARC AI2 (93.7%), HellaSwag (85.2%), BBH (83.3%).
Puntuaciones de benchmark
| Benchmark | Categor铆a | Puntuaci贸n | Bar |
|---|---|---|---|
| ARC AI2 | knowledge | 93.7 | |
| HellaSwag | knowledge | 85.2 | |
| BBH | reasoning | 83.3 | |
| MMLU | knowledge | 82.9 | |
| TriviaQA | knowledge | 82.9 | |
| Winogrande | knowledge | 70.4 | |
| PIQA | knowledge | 69.4 | |
| MATH level 5 | math | 64.8 | |
| Aider polyglot | coding | 48.4 | |
| GPQA diamond | knowledge | 42.0 | |
| OTIS Mock AIME 2024-2025 | math | 15.8 | |
| SimpleBench | reasoning | 2.7 | |
| FrontierMath-2025-02-28-Private | math | 1.7 |
Modelos similares
Google
56.9
Google
57.6
Mistral AI
57.8
Mistral
57.8