SecRepoBench (Standalone) - Secure-Pass@1: leaderboard

Metric: Completions that pass the unit tests and the PoC security test, pass@1 (%). Source: secrepobench.github.io. 29 models tracked.

Top models

#ModelScore
1GPT-5 (Medium)39.3
2O3 (Medium)32.4
3Claude Sonnet 4.5 (Thinking)31.1
4Claude Sonnet 4 (Thinking)30.2
5Claude 3.7 Sonnet (Thinking)28
6GPT-4.127.7
7Gemini 3 Pro27.4
8O4 Mini (Medium)24.5
9DeepSeek R123.9
10O1 (Medium)23.6
11Qwen3 Coder23
12DeepSeek V322.6
13GPT-4o (2024-11-20)22
14GPT-OSS-120B (Medium)21.4
15O3 Mini (Medium)21.4

Interactive version: theaggregate.ai/benchmark?slug=secrepobench-standalone-secure-pass-1 · How It Works · Data refreshed daily, snapshot 2026-09-19.