DeepSeek-V2 (MoE-236B, May 2024) vs phi-3-medium 14B
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | DeepSeek-V2 (MoE-236B, May 2024) | phi-3-medium 14B |
|---|---|---|
| Provider |  DeepSeek |  Microsoft |
| puntuación promedio | 76.5 | 58.6 |
| Precio de entrada | - | - |
| Precio de salida | - | - |
| Ventana de contexto | - | - |
| Publicado el | 2024-01-01 | 2024-01-01 |
| Código abierto | Open Source | Open Source |
Puntuaciones de benchmark
6 benchmarks · DeepSeek-V2 (MoE-236B, May 2024): 5, phi-3-medium 14B: 1
| Benchmark | Categoría | DeepSeek-V2 (MoE-236B, May 2024) | phi-3-medium 14B |
|---|---|---|---|
| ARC AI2 | knowledge | 89.6 | 88.8 |
| BBH | reasoning | 71.7 | 75.2 |
| HellaSwag | knowledge | 82.8 | 76.5 |
| MMLU | knowledge | 71.2 | 70.7 |
| TriviaQA | knowledge | 80.0 | 73.9 |
| Winogrande | knowledge | 72.6 | 63.0 |