Benchmark
Sayf SECURE CWET accuracy
Higher scores are better.
Benchmark sourceScores from Sayf cybersecurity evaluations data docs.
- Models measured
- 6
- Scores recorded
- 6
- Sources
- 1
- IQ weight
- 0.72
Score distribution
- 0.922
- Mean
- 0.923
- Median
- 0.023
- Std dev
- 0.9
- 10th percentile
- 0.95
- 90th percentile
- 0.889
- Min
This benchmark has a weight of 0.72 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 4.6secure_cwet· Sayf cybersecurity evaluations data docs | 0.95 |
| 2 | GPT 5.4secure_cwet· Sayf cybersecurity evaluations data docs | 0.95 |
| 3 | Gemma 4 31B ITsecure_cwet· Sayf cybersecurity evaluations data docs | 0.923 |
| 4 | Qwen3.6 35B A3Bsecure_cwet· Sayf cybersecurity evaluations data docs | 0.92 |
| 5 | Llama 3.3 70B Instructsecure_cwet· Sayf cybersecurity evaluations data docs | 0.9 |
| 6 | GPT OSS 20bsecure_cwet· Sayf cybersecurity evaluations data docs | 0.889 |
Data credits: Sayf-eval. Copyright (c) 2026 Qatar Computing Research Institute (QCRI). MIT License. (licence)