Specification Alignment on SPECBENCH
96.8Safety ScoreGPT-5.4
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GPT-5.4Model=GPT-5.42025.09 | 96.8 | 84.1 | 85.3 | |
| Gemini-3.1-proModel=Gemini-3.1-pro2025.09 | 84.7 | 75.6 | 72.1 | |
| Kimi-K2.5Model=Kimi-K2.52025.09 | 83.5 | 81.1 | 74.9 | |
| DeepSeek-V3.2Model=DeepSeek-V3.22025.09 | 77.6 | 79 | 66.6 |