Tool-use behavioral profiling on Malicious regime n=30 benchmark leaderboard · SOTA2 Research

Tool-use behavioral profiling on Malicious regime n=30

50Metric Amistral:latest

Evaluation Results

Method	Links
mistral:latest2026.04		50	40	10
mistral:latest2026.04		50	27	23
deepseek-chat2026.04		43	57	0
mistral:latest2026.04		43	23	33
qwen2.5:7b-instruct2026.04		30	63	20
qwen2.5:7b-instruct2026.04		20	60	20
deepseek-chat2026.04		10	90	0
llama3.1:latest2026.04		7	90	17
llama3.1:latest2026.04		7	93	7
qwen2.5:7b-instruct2026.04		7	97	3
llama3.1:latest2026.04		3	100	3
deepseek-chat2026.04		0	100	0