GPT-4 Turbo
来自 OpenAI · 发布于 2024-04-09
51.0
平均分
$10.00/1M
输入价格
$30.00/1M
输出价格
128K tokens (~64 books)
上下文窗口
multimodal
类型
Tested on 12 benchmarks with 51.0% average. Top scores: HellaSwag (93.7%), GSM8K (90.0%), TriviaQA (84.8%).
基准测试分数
| 基准测试 | 类别 | 分数 | Bar |
|---|---|---|---|
| HellaSwag | knowledge | 93.7 | |
| GSM8K | math | 90.0 | |
| TriviaQA | knowledge | 84.8 | |
| MMLU | knowledge | 76.5 | |
| Winogrande | knowledge | 75.0 | |
| CMMLU | knowledge | 71.0 | |
| BBH | reasoning | 66.8 | |
| MATH level 5 | math | 23.0 | |
| WeirdML | coding | 12.4 | |
| SimpleBench | reasoning | 10.1 | |
| GPQA diamond | knowledge | 7.5 | |
| OTIS Mock AIME 2024-2025 | math | 1.0 |
相似模型
OpenAI
51.1
DeepSeek
51.1
Google DeepMind
50.9
z-ai
50.8