Benchmark
Sayf CTI-Bench TAA accuracy
Higher scores are better.
Benchmark sourceScores from Sayf cybersecurity evaluations data docs.
- Models measured
- 6
- Scores recorded
- 6
- Sources
- 1
- IQ weight
- 0.71
Score distribution
- 0.477
- Mean
- 0.5
- Median
- 0.224
- Std dev
- 0.26
- 10th percentile
- 0.82
- 90th percentile
- 0.2
- Min
This benchmark has a weight of 0.71 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 4.6cti_taa· Sayf cybersecurity evaluations data docs | 0.82 |
| 2 | GPT 5.4cti_taa· Sayf cybersecurity evaluations data docs | 0.7 |
| 3 | Gemma 4 31B ITcti_taa· Sayf cybersecurity evaluations data docs | 0.5 |
| 4 | Llama 3.3 70B Instructcti_taa· Sayf cybersecurity evaluations data docs | 0.38 |
| 5 | GPT OSS 20bcti_taa· Sayf cybersecurity evaluations data docs | 0.26 |
| 6 | Qwen3.6 35B A3Bcti_taa· Sayf cybersecurity evaluations data docs | 0.2 |
Data credits: Sayf-eval. Copyright (c) 2026 Qatar Computing Research Institute (QCRI). MIT License. (licence)