CWE-bench v1: Java
Code · 2026-09-25
Programmatic pass@1 across the 20 Java tasks in CWE-bench v1, averaging four rollouts per task.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 68.0 |
| GPT-6 Astra | 65.0 |
| Claude Fable 5.1 | 60.0 |
| Grok 4.7 | 60.0 |
| GPT-6 Sol | 58.0 |
| Gemini 4 Argon | 56.0 |
| DeepSeek V4.1 Flash | 46.0 |
| Muse Spark 1.3 | 46.0 |
| Grok 4.6 | 41.0 |
| Hy4 Preview | 39.0 |
| Inkling | 33.0 |