CWE-bench v2 Public Set Pass@1
Cybersecurity · 2026-11-02
Pass@1 on the 100 tasks across 91 CWEs that Collinear AI publishes with the grader and harness, averaging four rollouts per task. Unlike the held-out set, these tasks can be seen and run by anyone.
Benchmark source: Cwe Bench
Top models (higher is better)
| Model | Score | Source |
|---|---|---|
| Gemini 4 Argon | 67.8 | CWE-bench v2 leaderboard (snapshot 2026-11-03) · Collinear AI |
| Claude Opus 5.5 | 67.3 | CWE-bench v2 leaderboard (snapshot 2026-11-03) · Collinear AI |
| Muse Wonder | 64.8 | CWE-bench v2 leaderboard (snapshot 2026-11-03) · Collinear AI |
| GPT-6 Astra | 63.8 | CWE-bench v2 leaderboard (snapshot 2026-11-03) · Collinear AI |
| DeepSeek V4.1 Flash 1017 | 57.0 | CWE-bench v2 leaderboard (snapshot 2026-11-03) · Collinear AI |
| Grok 4.7 | 55.5 | CWE-bench v2 leaderboard (snapshot 2026-11-03) · Collinear AI |
| Hy4 Preview | 52.0 | CWE-bench v2 leaderboard (snapshot 2026-11-03) · Collinear AI |
| Mistral Large 4 Preview | 46.3 | CWE-bench v2 leaderboard (snapshot 2026-11-03) · Collinear AI |
| Qwen3.8 Max (0902) | 45.0 | CWE-bench v2 leaderboard (snapshot 2026-11-03) · Collinear AI |
| GLM-5.3 | 44.0 | CWE-bench v2 leaderboard (snapshot 2026-11-03) · Collinear AI |
| Kimi K3.1 Pro | 41.5 | CWE-bench v2 leaderboard (snapshot 2026-11-03) · Collinear AI |