TheCluster/Qwen3.5-9B-Ultra-Heretic-MLX-mxfp8

Qwen3.5-9B Ultra Heretic

Quality: quantized (mxfp8, group size: 32, 8.626 bpw)

This is a abliterated (uncensored) version of Qwen/Qwen3.5-9B, made using Heretic v1.2.0 with Magnitude-Preserving Orthogonal Ablation (MPOA) and Self-Organizing Map Abliteration (SOMA)

Performance

Metric	This model	Original model (Qwen3.5-9B)
KL divergence	0.1085	0 (by definition)
Refusals	2/100	86/100

Lower refusals indicate fewer content restrictions, while lower KL divergence indicates better preservation of the original model's capabilities.

Alternative fine-tuned version: TheCluster/Qwen3.5-9B-Claude-4.6-HighIQ-INSTRUCT-HERETIC-UNCENSORED-MLX-mxfp8

Sampling Parameters:

• I suggest using the following sets of sampling parameters depending on the mode and task type:
  • Thinking mode for general tasks:
    temperature=1.0, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=1.5, repetition_penalty=1.0
  • Instruct (or non-thinking) mode for general tasks:
    temperature=0.7, top_p=0.8, top_k=20, min_p=0.0, presence_penalty=1.5, repetition_penalty=1.0
  • Instruct (or non-thinking) mode for reasoning tasks:
    temperature=1.0, top_p=1.0, top_k=40, min_p=0.0, presence_penalty=2.0, repetition_penalty=1.0
• For supported frameworks, you can adjust the presence_penalty parameter between 0 and 2 to reduce endless repetitions. However, using a higher value may occasionally result in language mixing and a slight decrease in model performance.

---

Source

This model was converted to MLX format from llmfan46/Qwen3.5-9B-ultra-heretic using mlx-vlm version 0.4.