Benchmark
Sayf SecEval accuracy
Higher scores are better.
Benchmark sourceScores from Sayf cybersecurity evaluations data docs.
- Models measured
- 6
- Scores recorded
- 6
- Sources
- 1
- IQ weight
- 0.89
Score distribution
- 0.748
- Mean
- 0.742
- Median
- 0.042
- Std dev
- 0.718
- 10th percentile
- 0.82
- 90th percentile
- 0.697
- Min
This benchmark has a weight of 0.89 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 | GPT 5.4seceval· Sayf cybersecurity evaluations data docs | 0.82 |
| 2 | Gemma 4 31B ITseceval· Sayf cybersecurity evaluations data docs | 0.783 |
| 3 | Qwen3.6 35B A3Bseceval· Sayf cybersecurity evaluations data docs | 0.742 |
| 4 | GPT OSS 20bseceval· Sayf cybersecurity evaluations data docs | 0.726 |
| 5 | Claude Sonnet 4.6seceval· Sayf cybersecurity evaluations data docs | 0.718 |
| 6 | Llama 3.3 70B Instructseceval· Sayf cybersecurity evaluations data docs | 0.697 |
Data credits: Sayf-eval. Copyright (c) 2026 Qatar Computing Research Institute (QCRI). MIT License. (licence)