# YOLOv8-N

> YOLOv8-N: one-stage object detector from the YOLOv8 family. Planned for LibreYOLO (not yet packaged).

- Family: YOLOv8 (External baseline)
- Parameters: 3.2M
- GFLOPs: 8.7
- Default input size: 640px
- Detection: anchor-free · nms
- Architecture: one-stage · CSPDarknet backbone
- Weight license: non-permissive
- Paper-reported mAP@50-95: 37.3%

## Benchmarks (COCO val2017)

| Hardware | Runtime | mAP@50-95 | FPS | Latency (ms) | VRAM (MB) |
|---|---|---|---|---|---|
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | 36.7 | 7.3 | 137.1 | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 36.7 | 26.4 | 37.8 | 34 |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 36.7 | 49.5 | 20.2 | 15 |
| Raspberry Pi 5 | HailoRT INT8 | 29.3 | 80.5 | 12.4 | - |
| Raspberry Pi 5 | ncnn FP32 | 36.7 | 11.9 | 83.7 | - |
| Raspberry Pi 5 | ONNX Runtime FP32 | 36.7 | 5.9 | 170.4 | - |
| Raspberry Pi 5 | PyTorch FP32 | 36.7 | 5.3 | 186.8 | - |


Source: https://www.visionanalysis.org/model/yolov8n. Benchmarks produced with LibreYOLO (https://github.com/Libre-YOLO/libreyolo).