# YOLO11-S

> YOLO11-S: one-stage object detector from the YOLO11 family. Planned for LibreYOLO (not yet packaged).

- Family: YOLO11 (External baseline)
- Parameters: 9.4M
- GFLOPs: 21.5
- Default input size: 640px
- Detection: anchor-free · nms
- Architecture: one-stage · CSPDarknet backbone
- Weight license: non-permissive
- Paper-reported mAP@50-95: 47%

## Benchmarks (COCO val2017)

| Hardware | Runtime | mAP@50-95 | FPS | Latency (ms) | VRAM (MB) |
|---|---|---|---|---|---|
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | 45.7 | 3.8 | 265.9 | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 45.7 | 20.7 | 48.3 | 110 |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 45.7 | 41.3 | 24.2 | 14 |
| Raspberry Pi 5 | ncnn FP32 | 45.7 | 5.7 | 174.5 | - |
| Raspberry Pi 5 | ONNX Runtime FP32 | 45.7 | 2.6 | 378.1 | - |
| Raspberry Pi 5 | PyTorch FP32 | 45.7 | 2.3 | 435.8 | - |


Source: https://www.visionanalysis.org/model/yolo11s. Benchmarks produced with LibreYOLO (https://github.com/Libre-YOLO/libreyolo).