// about
Who builds kqlbench
We measure where AI models actually land on natural-language-to-KQL, so security teams can make evidence-based decisions about adopting them in their detection workflows.
// mission
Close the gap between AI and security operations. Through rigorous, transparent benchmarking we show which models can genuinely support threat detection — and at what cost.
// team
// impact
Real-world testing
188 validated scenarios drawn from actual cybersecurity threats.
AI evaluation
Consistent testing across 14+ leading language models.
Open research
Transparent methodology and results for the security community.
// get in touch
Questions about the benchmark?
Connect with us on LinkedIn or X, or jump straight into the results.