Downloads · 30 days
76
13% of all-time downloads
YOLO26n-OptiQ-6bit
mlx-community/YOLO26n-OptiQ-6bit
YOLO26n-OptiQ-6bit is a object detection model from mlx-community. Use it when you need objects located in an image. It is set up for mlx. The card lists the license as agpl-3.0.
Mixed-precision quantized YOLO26n for Apple Silicon via optiq
All-time downloads
601
Public
Repo size
7.9 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors2.7 MB · 99%
At a glance
- Task
- Object Detection
- Library
- mlx
- License
- agpl-3.0
- Access
- Public
- Created
- Apr 10, 2026
- Updated
- Sep 28, 2026
- SHA
- 9f0fba8d
Base models
Model card
From the Hugging Face model README
YOLO26n-OptiQ-6bit
Mixed-precision quantized YOLO26n for Apple Silicon via optiq
This is a mixed-precision quantized version of YOLO26n in MLX format, optimized with mlx-optiq for Apple Silicon inference via yolo-mlx.
Quantization Details
| Property | Value |
|---|---|
| Target BPW | 6.0 |
| Achieved BPW | 5.96 |
| Layers at 4-bit | 12 |
| Layers at 8-bit | 114 |
| Original size | 9.9 MB |
| Quantized size | 2.5 MB |
| Compression | 3.9x |
Benchmark Results (COCO128)
| Model | Total Detections | Avg/Image |
|---|---|---|
| optiq 6-bit | 548 | 4.3 |
| Original (FP32) | 557 | 4.4 |
Detection delta: -9 (-1.6%) at 3.9x compression.
Usage
Requires mlx-optiq and yolo-mlx:
pip install mlx-optiq yolo-mlx
from optiq.models.yolo import load_quantized_yolo
model = load_quantized_yolo("mlx-community/YOLO26n-OptiQ-6bit")
results = model.predict("image.jpg")
How optiq Works
optiq measures each conv layer's sensitivity via KL divergence on detection outputs, then assigns optimal per-layer bit-widths using greedy knapsack optimization. Sensitive layers (detection head, feature pyramid) get 8-bit precision while robust backbone layers get 4-bit.
Credits
- Task
- Object Detection
- Library
- mlx
- License
- agpl-3.0
- Created
- Apr 10, 2026
- Updated
- Sep 28, 2026