CWE-bench v1: C/C++
Code · 2026-09-25
Programmatic pass@1 across the 31 C/C++ tasks in CWE-bench v1, averaging four rollouts per task.
Top models (higher is better)
| Model | Score |
|---|---|
| Grok 4.7 | 75.0 |
| GPT-6 Astra | 73.0 |
| Gemini 4 Argon | 72.0 |
| Claude Opus 5.5 | 69.0 |
| Claude Fable 5.1 | 65.0 |
| DeepSeek V4.1 Flash | 60.0 |
| GPT-6 Sol | 60.0 |
| Grok 4.6 | 60.0 |
| Hy4 Preview | 54.0 |
| Muse Spark 1.3 | 52.0 |
| Inkling | 27.0 |