XOR Vulnerability Repair - Exploit (OpenCode): leaderboard
XOR's automated vulnerability repair benchmark, attacker role: each system writes a working exploit for memory-safety vulnerabilities in low-level software with public fixes, checked by a withheld grader. This board holds the OpenCode harness runs from April 2026 (published September 2026); score is the share of vulnerabilities exploited, 0-100, higher means a more capable attacker. Reasoning effort is not recorded.
Metric: Pass Rate (%). Source: www.xor.tech. Saturation forecast: Around February 2027. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 96.2 | #33 |
| 2 | GPT-5.3 Codex | 77.3 | #72 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=xor-vulnerability-repair-exploit-opencode · How It Works · Data refreshed daily, snapshot 2026-11-05.