Cybersecurity Knowledge Evaluation on Cybermetric 2000

94.1AccuracyGPT-5

Evaluation results

Method
GPT-52026.01	94.1	0.002
GPT-4.12026.01	93.7	0.003
GPT-5-Mini2026.01	93.2	0.002
Llama-3.3-70B-Instruct2026.01	93	0.003
O3-Mini2026.01	93	0.003
GPT-OSS-120B2026.01	92.6	0.002
gpt-oss-120B2025.10	92.55	—
GPT-5-Nano2026.01	91.8	0.003
Phi-42026.01	91.2	0.004
Qwen-3-14B2026.01	91	0.002
Llama-Primus-Nemotron-70B-Instruct2026.01	90.6	0.003
Qwen-3-8B2026.01	90.3	0.002
gpt-oss-20B2025.10	90.2	—
CyberPal-2.0-14B2025.10	89.95	—
CyberPal-2.0-8B2025.10	89.85	—
Qwen3-14B2025.10	89.85	—
GPT-OSS-20B2026.01	89.3	0.005
CyberPal-2.0-20B2025.10	89.05	—
Qwen3-8B2025.10	88.45	—
CyberPal-2.0-4B2025.10	87.8	—
Qwen3-4B2025.10	87.4	—
Llama-Primus-Merged2026.01	86.2	0.002
Llama-3.1-8B-Instruct2026.01	85.1	0.004
Foundation-Sec-8B-Instruct2026.01	84.7	0.005
Foundation-Sec-8B-Reasoning2026.01	84.3	0.003
DeepHat-V1-7B2026.01	73	0.005