Share your thoughts, 1 month free Claude Pro on usSee more

Cybersecurity Evaluation on SecBench

88.8Accuracy

GPT-5

Updated 1mo ago

Evaluation Results

Method	Links
GPT-5 2026.01		88.8	0.4
GPT-5-Mini 2026.01		87.7	0.3
GPT-4.1 2026.01		87.2	0.2
O3-Mini 2026.01		86.9	0.5
GPT-OSS-120B 2026.01		85.3	0.4
Qwen-3-14B 2026.01		84.9	0.4
Llama-3.3-70B-Instruct 2026.01		84.2	0.5
GPT-5-Nano 2026.01		83.8	0.7
Qwen-3-8B 2026.01		83.5	0.7
Llama-Primus-Nemotron-70B-Instruct 2026.01		83	0.6
Phi-4 2026.01		81.3	0.4
GPT-OSS-20B 2026.01		80.4	0.5
Llama-3.1-8B-Instruct 2026.01		74.9	0.7
Llama-Primus-Merged 2026.01		74.9	0.4
Foundation-Sec-8B-Instruct 2026.01		74.4	1.4
Foundation-Sec-8B-Reasoning 2026.01		72.5	1.5
DeepHat-V1-7B 2026.01		61.5	0.7