Benchmark
Sayf RedSage: cybersecurity_tools_kali accuracy
Safety benchmarks examine risks in model behaviour, including how models respond to unsafe requests. Higher scores are better.
- Models measured
- 6
- Scores recorded
- 6
- Sources
- 1
- IQ weight
- 1.04
Score distribution
- 0.831
- Mean
- 0.833
- Median
- 0.028
- Std dev
- 0.805
- 10th percentile
- 0.872
- 90th percentile
- 0.803
- Min
This benchmark has a weight of 1.04 for chat models, measured over 6 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6redsage_kali· Sayf cybersecurity evaluations data docs | 0.872 |
| 2 | GPT 5.4redsage_kali· Sayf cybersecurity evaluations data docs | 0.864 |
| 3 | Qwen3.6 35B A3Bredsage_kali· Sayf cybersecurity evaluations data docs | 0.833 |
| 4 | Gemma 4 31B ITredsage_kali· Sayf cybersecurity evaluations data docs | 0.812 |
| 5 | GPT OSS 20bredsage_kali· Sayf cybersecurity evaluations data docs | 0.805 |
| 6 | Llama 3.3 70B Instructredsage_kali· Sayf cybersecurity evaluations data docs | 0.803 |
Data credits: Sayf-eval. Copyright (c) 2026 Qatar Computing Research Institute (QCRI). MIT License. (licence)