Secret Detection on LLMSec-DB benchmark leaderboard · SOTA2 Research

Secret Detection on LLMSec-DB

82.7EMAGPT-5

Evaluation Results

Method	Links
GPT-52025.10		82.7	91.7	50	3	0	42	16	2	38	9	4
Claude Sonnet 42025.10		81	87.7	50	0	0	44	10	0	39	8	3
Claude 3.7 Sonnet2025.10		80.3	92	50	0	0	43	21	1	37	12	4
Qwen-2.5 72B2025.10		80	91	50	2	0	44	11	0	38	11	3
Llama-3.1 405B2025.10		80	87.7	49	2	1	44	11	0	36	10	5
Qwen3 235B-A22b2025.10		78.7	86	48	3	2	42	2	2	33	8	8
Gemini 2.5 Flash Lite2025.10		77	86	45	1	5	44	8	0	33	10	8
GPT-5 Nano2025.10		76.7	85	50	6	0	44	17	0	34	13	7
Llama-3.3 70B2025.10		76	86.3	49	5	1	44	7	0	35	12	6
Gemini 2.5 Pro2025.10		75	93.3	50	0	0	42	24	2	32	14	8
Deepseek R12025.10		75	86.7	48	5	2	43	17	1	31	16	9
GPT-5-mini2025.10		71.3	84.3	50	11	0	44	24	0	30	17	10
NVIDIA Nemotron 9B2025.10		70.3	80.3	50	15	0	44	19	0	29	18	11
GPT-4o mini2025.10		67.7	86.3	50	13	0	43	9	1	28	17	13
Grok42025.10		67	81.7	49	0	1	42	20	2	27	20	14
Gemma 3 27B2025.10		65.7	82	50	18	0	44	25	0	27	19	15
Gemini Flash-1.5 8B2025.10		65	74.3	48	10	2	43	5	1	25	20	17
Mixtral 8x22b2025.10		63.3	77.3	43	3	7	44	25	0	22	24	20
Mistral 7B2025.10		59	76.7	46	11	4	40	13	4	17	29	25
Llama-4 Scout2025.10		59	80	50	43	0	44	20	0	16	30	26
Claude-3.5 Haiku2025.10		56	79	50	12	0	44	42	0	18	28	24
Gemma-3 12B2025.10		55.7	80	50	27	0	44	24	0	19	27	23
Qwen2.5 7B2025.10		54.7	65.7	47	26	3	36	24	8	20	26	22
Mixtral 8x7b2025.10		54	65.3	38	2	12	31	6	13	21	25	21
Ministral 3B2025.10		49.3	56.7	37	2	13	23	11	21	14	32	28