Huihui-Qwen3.8-27B-abliterated — SceneWorks MLX rehost (q8 / q4)
Pre-quantized MLX tiers of
huihui-ai/Huihui-Qwen3.8-27B-abliterated
at revision 739e3c5b89849f6c238ce1e5b70008612ae42cdd, an abliterated build of
Qwen/Qwen3.8-27B. Published by SceneWorks so the
tiers resolve by an immutable commit SHA for SceneWorks Inference's MLX language-model engine,
and so a q4 user does not download the 55.6 GB full-precision files. It is not an official
Qwen or huihui-ai distribution.
Layout
| Dir | Contents |
|---|---|
q8/ |
pre-quantized Q8 tier (30 GB): one model.safetensors, config, tokenizer, chat template |
q4/ |
pre-quantized Q4 tier (19.98 GB): one model.safetensors, config, tokenizer, chat template |
Each tier directory is self-contained. Download only the tier you use. Full precision is not mirrored here; take it from the upstream repository.
The tiers are produced from the upstream bf16 snapshot by mlx-llm's snapshot preparer
(prepare_snapshot). The decoder and MTP projections are stored as MLX group-wise affine
quantized weights (group size 64); config.json carries quantization: {bits, group_size}.
The vision tower (model.visual.*), embeddings and norms stay bf16 in every tier, so the model
still accepts image input.
Provenance
SOURCE_REVISION.json records the upstream repository and revision, the base model, and the
SceneWorks Inference revision that produced the tiers.
License and use
Apache-2.0, as released by Qwen and kept by huihui-ai; see LICENSE. The upstream model card
states that abliteration removes the model's refusal behaviour and that its output may be
inappropriate for some audiences. That applies to these tiers unchanged. You are responsible
for how you use it.
Quantized
Model tree for SceneWorks/qwen3.8-27b-abliterated-mlx
Base model
Qwen/Qwen3.8-27B