Benchmark
CTF Speedrun
A model benchmark based on CTF (Capture The Flag) cybersecurity challenges, measuring a model's ability to independently complete cybersecurity challenges within a limited time. The benchmark covers multiple categories of security techniques, including reverse engineering, exploit development, and cryptography.
Timeline (last 90 days)
Oct 7
In CTF Speedrun (a 19-challenge cybersecurity speed competition), Mistral Large 4 was the fastest model at solving challenges and solved 18 out of 19 problems.
Unverified50%