BinaryAudit: leaderboard
BinaryAudit (Quesma) — auditing compiled binaries for vulnerabilities and behavior without source access; average success rate across audit tasks.
Metric: Avg Success Rate (%). Source: quesma.com. Saturation forecast: Not forecast. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.3 Codex | 94.12 | #69 |
| 2 | GPT-5.2 Codex (xHigh) | 90.67 | #95 (GPT-5.2 Codex) |
| 3 | GPT-5 | 88.89 | #98 |
| 4 | GPT-5.2 Codex (High) | 88.41 | #95 (GPT-5.2 Codex) |
| 5 | Grok 4.1 Fast | 86.11 | #217 |
| 6 | Gemini 3.1 Pro (Preview) | 84.62 | #58 |
| 7 | GPT-5.4 | 84.44 | #79 |
| 8 | DeepSeek V3.2 | 83.33 | #205 |
| 9 | GPT-5.3 Codex (High) | 82.46 | #69 (GPT-5.3 Codex) |
| 10 | Claude Opus 4.6 | 82.05 | #63 |
| 11 | GPT-5.4 (High) | 81.67 | #79 (GPT-5.4) |
| 12 | GPT-5.3 Codex (xHigh) | 79.49 | #69 (GPT-5.3 Codex) |
| 13 | Claude Sonnet 4.5 | 79.37 | #142 |
| 14 | Claude Opus 4.5 | 76 | #83 |
| 15 | Kimi K2.5 | 75.93 | #144 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=binaryaudit · How It Works · Data refreshed daily, snapshot 2026-10-20.