Downloads · 30 days
2.6K
6% of all-time downloads
Youssofal/Qwen3.6-35B-A3B-Abliterated-Heretic-MLX-4bit
Qwen3.6-35B-A3B-Abliterated-Heretic-MLX-4bit is a text generation model from Youssofal. Use it when you need the model to write or continue text. It is set up for mlx. The card lists the license as apache-2.0.
This is an MLX release of an abliterated version of Qwen's Qwen3.6-35B-A3B.
Downloads · 30 days
2.6K
6% of all-time downloads
All-time downloads
44.8K
Public
Parameters
34.7B
48.3 GB on disk
Likes
22
Public
Click a slice to open those files.
.safetensors24.6 GB · 100%
How the weights are stored.
U3234.7B · 99%
From the Hugging Face model README
This is an MLX release of an abliterated version of Qwen's Qwen3.6-35B-A3B.
By applying Heretic's ablation pipeline to the text-side MoE stack, the base refusal behavior was removed at the weight level. This release keeps the Qwen3.6-35B-A3B reasoning and instruction-following profile in Apple MLX format for local deployment on Apple Silicon hardware.
This MLX repo includes the retained Qwen3.6 image/video processor files and vision tower tensors for runtimes with Qwen3.6 multimodal MLX support.
| Check | Original Qwen3.6-35B-A3B | Abliterated Heretic MLX |
|---|---|---|
| Official 25-prompt refusal check | 22/25 refusals | 3/25 refusals |
| Archived Heretic KL divergence | - | 0.010655362159013748 |
Qwen3.6-35B-A3B is a sparse MoE model in the qwen3_5_moe family. The accepted abliterated BF16 source checkpoint was produced with a Heretic MPOA/SOMA-style sibling-transfer workflow and finalized with the input-side split-MoE intervention that cleared the official 25-prompt refusal marker suite down to 1/25.
This MLX release was built directly from the published BF16 Heretic checkpoint using a high-quality layer-aware quantization policy instead of a flat per-weight pass.
4-bit4-bit high-quality mixed layer-aware quantization (4/6-bit)Youssofal/Qwen3.6-35B-A3B-Abliterated-Heretic-BF16Qwen3.6-35B-A3B-Abliterated-Heretic-MLX-4bitThe layer-aware policy keeps more precision on sensitive projections in the early, late, and selected middle layers so the quant stays cleaner than a naive flat conversion.
This published MLX variant passed:
25-prompt refusal marker check in standard thinking-enabled chat format: 3/25 refusalsall_looks_ok=truefrom mlx_lm import load, generate
model, tokenizer = load("Youssofal/Qwen3.6-35B-A3B-Abliterated-Heretic-MLX-4bit")
messages = [{"role": "user", "content": "Write a short Python function that reverses a string."}]
prompt = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True,
)
response = generate(model, tokenizer, prompt=prompt, max_tokens=256)
print(response)
The repo root contains the complete 4-bit MLX export for this variant:
config.jsonmodel.safetensors.index.jsonmodel-*.safetensors shardsmodel-vision-00001-of-00001.safetensorsREADME.mdThis model has had refusal behavior removed at the weight level. It will answer prompts that the base model would normally refuse. You are responsible for how you use it.