SecRepoBench (Standalone) - Pass@1: leaderboard
Metric: Completions that pass the unit tests, pass@1 (%). Source: secrepobench.github.io. 29 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 (Medium) | 54.1 |
| 2 | Claude Sonnet 4.5 (Thinking) | 52.2 |
| 3 | Claude Sonnet 4 (Thinking) | 48.4 |
| 4 | Gemini 3 Pro | 48.1 |
| 5 | O3 (Medium) | 47.5 |
| 6 | GPT-4.1 | 43.4 |
| 7 | Qwen3 Coder | 41.2 |
| 8 | Claude 3.7 Sonnet (Thinking) | 40.3 |
| 9 | DeepSeek V3 | 39.6 |
| 10 | O1 (Medium) | 38.4 |
| 11 | O4 Mini (Medium) | 36.8 |
| 12 | Claude 3.5 Sonnet | 36.5 |
| 13 | GPT-4o (2024-11-20) | 34.9 |
| 14 | DeepSeek R1 | 34.3 |
| 15 | GPT-OSS-120B (Medium) | 34 |
Interactive version: theaggregate.ai/benchmark?slug=secrepobench-standalone-pass-1 · How It Works · Data refreshed daily, snapshot 2026-09-19.