Downloads · 30 days
31
8% of all-time downloads
Qwen3.8-27B-Quality
hanxin2000/Qwen3.8-27B-Quality
Qwen3.8-27B-Quality is a text generation model from hanxin2000. 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.
GDN-tiered MLX hybrid of Qwen/Qwen3.8-27B for oMLX. Built with mlxlm.convert plus re-embedded BF16 MTP. Not an oQ export.
All-time downloads
368
Public
Parameters
26.9B
26.3 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors26.3 GB · 100%
Parameter types
How the weights are stored.
U3225.4B · 93%
Base models
Model card
From the Hugging Face model README
Qwen3.8-27B-Quality
GDN-tiered MLX hybrid of Qwen/Qwen3.8-27B for oMLX. Built with mlx_lm.convert plus re-embedded BF16 MTP. Not an oQ export.
About 26 GB on disk. Intended as a quality-leaning local checkpoint: keep GDN gates and MTP in BF16, protect attention and residual down_proj, compress the bulky MLP gate/up path.
Quantization
| Precision | Tensors |
|---|---|
| BF16 | Entire MTP head; per-layer GDN A_log, dt_bias, conv1d, in_proj_a, in_proj_b, in_proj_z |
| 8-bit affine / gs64 | embed_tokens, lm_head, all self_attn, linear_attn.in_proj_qkv, linear_attn.out_proj, MLP on layers 0–7 and 56–63 |
| 6-bit affine / gs32 | Middle-layer mlp.down_proj |
| 4-bit affine / gs32 | Middle-layer mlp.gate_proj and mlp.up_proj |
Lightning MTP stays attached as language_model.mtp.* in model-mtp.safetensors. mlx_lm.convert strips mtp.*; those 15 tensors are copied back from the BF16 source.
Recommended sampling
Qwen3.8 official defaults:
temperature=1.0top_p=0.95top_k=20- thinking on (can be disabled per request)
oMLX: enable Lightning MTP, mtp_num_draft_tokens=3.
Use in oMLX
Place the folder under ~/.omlx/models/Qwen3.8-27B-Quality (or Import). Enable MTP in model settings. Do not run oQ on this checkpoint if you want the BF16 MTP/gates kept.
Use with mlx-lm
from mlx_lm import load, generate
model, tokenizer = load("hanxin2000/Qwen3.8-27B-Quality")
print(generate(model, tokenizer, prompt="Hello", max_tokens=64))
MTP speculative decoding is an oMLX path. Stock mlx_lm.generate will run the quantized trunk; it may ignore the embedded MTP head.
Files
model-0000n-of-00005.safetensors— quantized language trunkmodel-mtp.safetensors— BF16 MTPmodel.safetensors.index.jsonconfig.json, tokenizer,chat_template.jinja
License
Apache 2.0, same as the Qwen3.8-27B source weights.
Acknowledgements
Base model: Qwen Team, Qwen3.8-27B. Runtime: oMLX / MLX.
- Task
- Text Generation
- Library
- mlx
- Type
- qwen3_5
- License
- apache-2.0
- Languages
- zh, en
- Created
- Aug 16, 2026
- Updated
- Aug 16, 2026