RedSage-Bench: leaderboard
Cybersecurity benchmark with 30K multiple-choice and 240 open-ended QA items covering knowledge, offensive skills, and tool expertise.
Metric: Macro Accuracy (imported). Source: risys-lab.github.io. Source: benchmarklist.com. Saturation forecast: Around October 2026. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 88.68 | #98 |
| 2 | Qwen 3 32B | 85.4 | #443 |
| 3 | Qwen3-8B-Base | 84.24 | #626 |
| 4 | Qwen 3 8B | 81.85 | #674 |
| 5 | Llama 3.1 8B | 78.02 | #1151 |
| 6 | Llama 3.1 8B Instruct | 77.05 | #1035 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=redsage-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.