AthenaBench - RCM: leaderboard
Map a vulnerability description to its CWE root cause, 2,000 items; exact-match accuracy.
Metric: Accuracy (%) mapping vulnerability descriptions to their CWE root cause (2,000 items). Source: raw.githubusercontent.com. Saturation forecast: Around February 2027. 10 models tracked. Updated 17 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-5 | 71.6 | #177 | |
| 2 | Gemini 2.5 Pro | 71.2 | #274 | |
| 3 | Gemini 2.5 Flash | 65.1 | #467 | |
| 4 | Llama 3.3 70B Instruct | 60 | #1387 | |
| 5 | Llama 3 70B Instruct | 56.7 | #1689 | |
| 6 | Llama-Primus-Merged | 56.6 | #2461 | |
| 7 | Qwen 3 14B | Reasoning | 54.1 | #1264 |
| 8 | Qwen 3 8B | Thinking | 48.9 | #1868 |
| 9 | Qwen 3 4B 2507 Instruct | 45.4 | #2215 | |
| 10 | Llama 3.1 8B Instruct | 42.8 | #4576 |
No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=athenabench-rcm · How It Works · Data refreshed daily, snapshot 2026-11-17.