Downloads · 30 days
52
26% of all-time downloads
NeoHorse-1-4B-oQ4e
0xPatronum/NeoHorse-1-4B-oQ4e
NeoHorse-1-4B-oQ4e is a text generation model from 0xPatronum. 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.
Native oMLX enhanced 4-bit quant of TokenRhythm/NeoHorse-1-4B at revision 56f0584bb40578a2c33b1b40a08ccd17243ad710.
All-time downloads
202
Public
Parameters
4.2B
2.5 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors2.5 GB · 99%
Parameter types
How the weights are stored.
U324.2B · 100%
Try a prompt
Base models
Model card
From the Hugging Face model README
<!-- vale Google.Units = NO -->
NeoHorse-1-4B oQ4e
Native oMLX enhanced 4-bit quant of
TokenRhythm/NeoHorse-1-4B
at revision 56f0584bb40578a2c33b1b40a08ccd17243ad710.
This pack is oMLX oQ4e, not an mlx-optiq OptiQ pack.
Quantization
| Setting | Value |
|---|---|
| Converter | oMLX quantize_oq_streaming |
| Level | oq_level=4, enhanced |
| Group size | 64 |
| Text only | yes |
| Calibration | built-in oqe_code_multilingual, 128 samples of length 512 |
| Imatrix entries | 248, none missing |
Source was the verified BF16 snapshot of that revision.
Use
Load the directory with oMLX. Mixed per-layer bit widths are part of oQ4e.
License
Apache License 2.0.
Derivative of TokenRhythm/NeoHorse-1-4B, post-trained from
Qwen/Qwen3.5-4B. Original model: Copyright 2026 Alibaba Cloud.
Quantization keeps the same license.
- Task
- Text Generation
- Library
- mlx
- Type
- qwen3_5
- License
- apache-2.0
- Created
- Sep 22, 2026
- Updated
- Sep 22, 2026