Model Zoo

Pre-trained weights for various YOLO architectures. Check the license badge before integrating into commercial products.

Understanding Weight Licenses

The Libre-YOLO engine is always MIT-licensed. However, pre-trained weights may have different origins:

MIT Weights

Trained from scratch using MIT-licensed code and public datasets. Safe for any commercial use.

Converted

Converted from other implementations. Original training may involve AGPL code. Review your use case.

8

Total Models

5

MIT Weights

2

Architectures

COCO

Dataset

MIT Licensed Weights

COMMERCIAL SAFE

YOLOv11n

Nano

MIT Weights

Ultra-lightweight for edge deployment

Params

2.6M

mAP@50-95

39.5

Speed (T4)

1.5ms

Download

YOLOv11s

Small

MIT Weights

Balanced performance and accuracy

Params

9.4M

mAP@50-95

47.0

Speed (T4)

2.5ms

Download

YOLOv11m

Medium

MIT Weights

Production workhorse

Params

20.1M

mAP@50-95

51.5

Speed (T4)

4.7ms

Download

YOLOv8n

Nano

MIT Weights

v8 architecture, edge-optimized

Params

3.2M

mAP@50-95

37.3

Speed (T4)

1.8ms

Download

YOLOv8s

Small

MIT Weights

v8 architecture, balanced

Params

11.2M

mAP@50-95

44.9

Speed (T4)

3.1ms

Download

Converted Weights

REVIEW REQUIRED

Note: These weights were converted from other implementations. While the Libre-YOLO engine is MIT-licensed, the original training pipeline may have involved AGPL-licensed code. Consult your legal team for commercial deployments using these weights.

YOLOv11l

Large

Converted Weights

High accuracy applications

Params

25.3M

mAP@50-95

53.4

Speed (T4)

6.2ms

Download

YOLOv11x

Extra Large

Converted Weights

Maximum accuracy

Params

56.9M

mAP@50-95

54.7

Speed (T4)

11.3ms

Download

YOLOv8m

Medium

Converted Weights

v8 architecture, high performance

Params

25.9M

mAP@50-95

50.2

Speed (T4)

5.8ms

Download

Need help choosing the right model for your use case?

Read the Commercial Guide