phi-3-medium 14B vs phi-3-mini 3.8B
并排对比,每项指标,每项基准测试。
| 类型 | phi-3-medium 14B | phi-3-mini 3.8B |
|---|---|---|
| Provider | ||
| 平均分 | 58.6 | 61.0 |
| 输入价格 | - | - |
| 输出价格 | - | - |
| 上下文窗口 | - | - |
| 发布于 | 2024-01-01 | 2024-01-01 |
| 开源 | Open Source | Open Source |
基准测试分数
8 benchmarks · phi-3-medium 14B: 7, phi-3-mini 3.8B: 1
| 基准测试 | 类别 | phi-3-medium 14B | phi-3-mini 3.8B |
|---|---|---|---|
| ANLI | knowledge | 33.7 | 29.2 |
| ARC AI2 | knowledge | 88.8 | 79.9 |
| BBH | reasoning | 75.2 | 62.3 |
| HellaSwag | knowledge | 76.5 | 68.9 |
| MMLU | knowledge | 70.7 | 58.4 |
| OpenBookQA | knowledge | 83.2 | 84.0 |
| TriviaQA | knowledge | 73.9 | 64.0 |
| Winogrande | knowledge | 63.0 | 41.6 |