Benchmark
Sayf SecBench MCQ (English) accuracy
Higher scores are better.
Benchmark sourceScores from Sayf cybersecurity evaluations data docs.
- Models measured
- 6
- Scores recorded
- 6
- Sources
- 1
- IQ weight
- 0.95
Score distribution
- 0.861
- Mean
- 0.877
- Median
- 0.032
- Std dev
- 0.821
- 10th percentile
- 0.897
- 90th percentile
- 0.816
- Min
This benchmark has a weight of 0.95 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 4.6secbench· Sayf cybersecurity evaluations data docs | 0.897 |
| 2 | Qwen3.6 35B A3Bsecbench· Sayf cybersecurity evaluations data docs | 0.891 |
| 3 | GPT 5.4secbench· Sayf cybersecurity evaluations data docs | 0.877 |
| 4 | Gemma 4 31B ITsecbench· Sayf cybersecurity evaluations data docs | 0.867 |
| 5 | Llama 3.3 70B Instructsecbench· Sayf cybersecurity evaluations data docs | 0.821 |
| 6 | GPT OSS 20bsecbench· Sayf cybersecurity evaluations data docs | 0.816 |
Data credits: Sayf-eval. Copyright (c) 2026 Qatar Computing Research Institute (QCRI). MIT License. (licence)