---
base_model: Goekdeniz-Guelmez/Qwen3-4B-Sky-High-Hermes-gabliterated
tags:
- uncensored
- gabliteration
- mlx
datasets:
- mlabonne/harmless_alpaca
- mlabonne/harmful_behaviors
library_name: mlx
arxiv: '2512.18901'
pipeline_tag: text-generation
model-index:
- name: ZeroXClem_Qwen3-4B-Sky-High-Hermes-gabliterated
  results:
  - task:
      type: text-generation
    dataset:
      name: Harmless Alpaca
      type: harmless_alpaca
    metrics:
    - type: pass@1
      value: 0.0992
      name: KL Divergence
  - task:
      type: text-generation
    dataset:
      name: Harmful Behaviors
      type: harmful_behaviors
    metrics:
    - type: pass@1
      value: 0.02
      name: Refusal Rate
---

# mlx-community/Qwen3-4B-Sky-High-Hermes-gabliterated-8bit

This model [mlx-community/Qwen3-4B-Sky-High-Hermes-gabliterated-8bit](https://huggingface.co/mlx-community/Qwen3-4B-Sky-High-Hermes-gabliterated-8bit) was
converted to MLX format from [Goekdeniz-Guelmez/Qwen3-4B-Sky-High-Hermes-gabliterated](https://huggingface.co/Goekdeniz-Guelmez/Qwen3-4B-Sky-High-Hermes-gabliterated)
using mlx-lm version **0.30.4**.

## Use with mlx

```bash
pip install mlx-lm
```

```python
from mlx_lm import load, generate

model, tokenizer = load("mlx-community/Qwen3-4B-Sky-High-Hermes-gabliterated-8bit")

prompt = "hello"

if tokenizer.chat_template is not None:
    messages = [{"role": "user", "content": prompt}]
    prompt = tokenizer.apply_chat_template(
        messages, add_generation_prompt=True, return_dict=False,
    )

response = generate(model, tokenizer, prompt=prompt, verbose=True)
```