Benchmark
Sayf CTI-Bench ATE (ATT&CK) recall
Higher scores are better.
Benchmark sourceScores from Sayf cybersecurity evaluations data docs.
- Models measured
- 6
- Scores recorded
- 6
- Sources
- 1
- IQ weight
- 1.49
Score distribution
- 0.645
- Mean
- 0.738
- Median
- 0.178
- Std dev
- 0.509
- 10th percentile
- 0.819
- 90th percentile
- 0.315
- Min
This benchmark has a weight of 1.49 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 4.6ate· Sayf cybersecurity evaluations data docs | 0.819 |
| 2 | GPT 5.4ate· Sayf cybersecurity evaluations data docs | 0.791 |
| 3 | Gemma 4 31B ITate· Sayf cybersecurity evaluations data docs | 0.738 |
| 4 | Qwen3.6 35B A3Bate· Sayf cybersecurity evaluations data docs | 0.7 |
| 5 | Llama 3.3 70B Instructate· Sayf cybersecurity evaluations data docs | 0.509 |
| 6 | GPT OSS 20bate· Sayf cybersecurity evaluations data docs | 0.315 |
Data credits: Sayf-eval. Copyright (c) 2026 Qatar Computing Research Institute (QCRI). MIT License. (licence)