GPT-5.2
来自 OpenAI · 发布于 2025-12-10
54.0
平均分
$1.75/1M
输入价格
$14.00/1M
输出价格
400K tokens (~200 books)
上下文窗口
multimodal
类型
Tested on 20 benchmarks with 54.0% average. Top scores: Chatbot Arena Elo — Overall (1439.5%), Chatbot Arena Elo — Coding (1403.1%), OTIS Mock AIME 2024-2025 (96.1%).
基准测试分数
| 基准测试 | 类别 | 分数 | Bar |
|---|---|---|---|
| Chatbot Arena Elo — Overall | arena | 1439.5 | |
| Chatbot Arena Elo — Coding | arena | 1403.1 | |
| OTIS Mock AIME 2024-2025 | math | 96.1 | |
| GPQA diamond | knowledge | 88.5 | |
| ARC-AGI | reasoning | 86.2 | |
| VPCT | knowledge | 76.0 | |
| SWE-Bench verified | coding | 73.8 | |
| WeirdML | coding | 72.2 | |
| SWE-Bench Verified (Bash Only) | coding | 71.8 | |
| Terminal Bench | coding | 64.9 | |
| ARC-AGI-2 | reasoning | 52.9 | |
| Chess Puzzles | knowledge | 49.0 | |
| FrontierMath-2025-02-28-Private | math | 40.7 | |
| SimpleQA Verified | knowledge | 38.9 | |
| SimpleBench | reasoning | 35.0 | |
| APEX-Agents | agentic | 34.3 | |
| GSO-Bench | coding | 27.4 | |
| HLE | knowledge | 24.2 | |
| PostTrainBench | knowledge | 21.4 | |
| FrontierMath-Tier-4-2025-07-01-Private | math | 18.8 |