Downloads · 30 days
268
30% of all-time downloads
Qwen3.8-27B-2bit
agnosticeng/Qwen3.8-27B-2bit
Qwen3.8-27B-2bit is a text generation model from agnosticeng. 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.
2-bit MLX quantization of Qwen/Qwen3.8-27B, with the Qwen3.8 MTP head bundled at mtp/ for speculative decoding.
All-time downloads
901
Public
Parameters
26.9B
8.8 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors8.6 GB · 99%
Parameter types
How the weights are stored.
U3226.9B · 100%
Try a prompt
Base models
Model card
From the Hugging Face model README
agnosticeng/Qwen3.8-27B-2bit
2-bit MLX quantization of Qwen/Qwen3.8-27B, with the Qwen3.8 MTP head bundled at mtp/ for speculative decoding.
Quantization
- Affine 2-bit, group size 64 (~2.7 bits/weight, ~8.4 GB)
- Converted with mlx-vlm; vision tower removed (text-only)
MTP head
mtp/weights.safetensors contains the Qwen3.8-27B multi-token-prediction sidecar (2-bit, mtp.-prefixed keys).
- Task
- Text Generation
- Library
- mlx
- Type
- qwen3_5
- License
- apache-2.0
- Created
- Sep 6, 2026
- Updated
- Sep 6, 2026