45 5 days ago

A Havenlon-focused Qwen3.8 27B model for execution-boundary reasoning, Adversarial Completeness, independent veto, evidence chains, and AI Agent execution control.

{
"num_ctx": 32768,
"repeat_penalty": 1.08,
"stop": [
"<|im_end|>"
],
"temperature": 0,
"top_p": 1
}