GPT-3.5 Turbo (older v0613)
来自 OpenAI · 发布于 2024-01-25
45.8
平均分
$1.00/1M
输入价格
$2.00/1M
输出价格
4K tokens (~2 books)
上下文窗口
text
类型
Tested on 13 benchmarks with 45.8% average. Top scores: TriviaQA (85.8%), ARC AI2 (83.2%), OpenBookQA (81.3%).
基准测试分数
| 基准测试 | 类别 | 分数 | Bar |
|---|---|---|---|
| TriviaQA | knowledge | 85.8 | |
| ARC AI2 | knowledge | 83.2 | |
| OpenBookQA | knowledge | 81.3 | |
| Winogrande | knowledge | 63.2 | |
| GSM8K | math | 57.8 | |
| MMLU | knowledge | 56.4 | |
| Aider — Code Editing | coding | 50.4 | |
| BBH | reasoning | 48.8 | |
| ANLI | knowledge | 37.1 | |
| CSQA2 | knowledge | 14.0 | |
| MATH level 5 | math | 11.6 | |
| WeirdML | coding | 3.5 | |
| GPQA diamond | knowledge | 2.9 |