DeepsecBench-AA: leaderboard
Artificial Analysis's private DeepsecBench run: agents statically review open-source application code pinned before major vulnerability fixes and report findings, which a GPT-5.6 Sol (high) judge validates and matches to an expert-verified golden set; F2 weights recall over precision. Median of three runs on AA's Stirrup harness; cells for models that declined more than 5% of tasks are withheld. F2 (%).
Metric: Median F2 (%). Source: artificialanalysis.ai. Saturation forecast: Around September 2028. 23 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Sol (Max) | 46.32 | #19 (GPT-6 Sol) |
| 2 | GPT-6 Astra (Max) | 36.93 | #6 (GPT-6 Astra) |
| 3 | Kimi K3.1 Pro | 36.71 | #17 |
| 4 | DeepSeek V4.1 Flash (Max) | 31.2 | #37 (DeepSeek V4.1 Flash) |
| 5 | Grok 4.7 (xHigh) | 26.92 | #26 (Grok 4.7) |
| 6 | MiMo-V2.6-Pro | 26.24 | #48 |
| 7 | Claude Opus 5 (Max) | 25.85 | #16 (Claude Opus 5) |
| 8 | DeepSeek V4.1 Flash (1017) (Max) | 24.93 | #27 (DeepSeek V4.1 Flash (1017)) |
| 9 | GLM-5.3 (Max) | 24.29 | #44 (GLM-5.3) |
| 10 | GPT-6 Luna (Max) | 23.53 | #61 (GPT-6 Luna) |
| 11 | GLM-5.3 Flash | 19.89 | #77 |
| 12 | Mistral Large 4 | 16.02 | #106 |
| 13 | Step 5 Preview | 15.81 | #53 |
| 14 | mistral-large-4-oss (High) | 14.96 | #60 (mistral-large-4-oss) |
| 15 | Gemini 3.8 Flash (High) | 14.54 | #30 (Gemini 3.8 Flash) |
No result here: #3 Claude Opus 5.5, #7 GPT-6.1 Sol, #9 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=deepsecbench-aa · How It Works · Data refreshed daily, snapshot 2026-11-05.