SAST False-Positive Triage: leaderboard
A static analyser fires 142 findings over a codebase and 11 of them are real. The model has to keep the genuine vulnerabilities and dismiss the noise, so the accuracy column punishes both misses and credulity. Published as a blog table, which is why the roster is small and current.
Metric: Accuracy (%). Source: www.fencer.dev. Saturation forecast: Around January 2027. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.8 | 92 | #37 |
| 2 | Kimi K3 | 92 | #30 |
| 3 | Claude Sonnet 5 | 89 | #57 |
| 4 | Claude Opus 5 | 88 | #12 |
| 5 | GPT-5.6 Luna | 82 | #68 |
| 6 | Claude Haiku 4.5 | 81 | #280 |
| 7 | GPT-OSS-120B | 80 | #341 |
| 8 | Devstral 2 | 78 | #411 |
| 9 | MiniMax-M2.5 | 76 | #303 |
| 10 | GPT-5.6 Terra | 75 | #41 |
| 11 | Kimi K2.5 | 73 | #144 |
| 12 | GPT-5.6 Sol | 73 | #19 |
| 13 | GLM-5 | 72 | #141 |
| 14 | Nemotron 3 Super | 71 | #335 |
| 15 | DeepSeek V3.2 | 63 | #205 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=sast-false-positive-triage · How It Works · Data refreshed daily, snapshot 2026-10-20.