AthenaBench - RMS: leaderboard
Propose ATT&CK mitigations for a described attack, 500 items; mean F1 of the mitigation IDs against the reference set.
Metric: F1 (%) of proposed ATT&CK mitigations against the reference set (500 items). Source: raw.githubusercontent.com. Saturation forecast: Around March 2027. 9 models tracked. Updated 17 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemini 2.5 Pro | 28.4 | #274 | |
| 2 | Gemini 2.5 Flash | 13.4 | #467 | |
| 3 | Llama 3.3 70B Instruct | 11.1 | #1387 | |
| 4 | Llama 3 70B Instruct | 11.1 | #1689 | |
| 5 | Qwen 3 14B | Reasoning | 7 | #1264 |
| 6 | Llama-Primus-Merged | 6.6 | #2461 | |
| 7 | Qwen 3 8B | Thinking | 5.5 | #1868 |
| 8 | Qwen 3 4B 2507 Instruct | 4.8 | #2215 | |
| 9 | Llama 3.1 8B Instruct | 3.6 | #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-rms · How It Works · Data refreshed daily, snapshot 2026-11-17.