# YOLOv8-L

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

- Family: YOLOv8 (External baseline)
- Parameters: 43.7M
- GFLOPs: 165.2
- Default input size: 640px
- Detection: anchor-free · nms
- Architecture: one-stage · CSPDarknet backbone
- Weight license: non-permissive
- Paper-reported mAP@50-95: 52.9%

## Benchmarks (COCO val2017)

| Hardware | Runtime | mAP@50-95 | FPS | Latency (ms) | VRAM (MB) |
|---|---|---|---|---|---|
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | 52.0 | 0.7 | 1451.2 | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 52.0 | 8.3 | 120.3 | 262 |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 52.0 | 19.3 | 51.9 | 14 |
| Raspberry Pi 5 | HailoRT INT8 | 44.3 | 17.2 | 58.1 | - |
| Raspberry Pi 5 | ncnn FP32 | 52.0 | 1.5 | 685.9 | - |
| Raspberry Pi 5 | ONNX Runtime FP32 | 52.0 | 0.5 | 1913.7 | - |
| Raspberry Pi 5 | PyTorch FP32 | 52.0 | 0.6 | 1762.9 | - |


Source: https://www.visionanalysis.org/model/yolov8l. Benchmarks produced with LibreYOLO (https://github.com/Libre-YOLO/libreyolo).