Back to Leaderboard
Parameters
29.5M
FLOPs
160.4G
Input Size
640px
License
MIT
Architecture
Type
one-stage
Backbone
CSPDarknet
Neck
PAFPN
Head
NMS-Free
Benchmark Results
Performance on COCO val2017 across different hardware configurations
| Hardware | mAP@50-95 | FPS | Latency | VRAM |
|---|---|---|---|---|
| NVIDIA A100 (TensorRT FP16) | 54.3% | 92.6 | 10.8ms | 1478 MB |
| NVIDIA T4 (TensorRT FP16) | 54.5% | 33.5 | 29.9ms | 1416 MB |
| CPU (ONNX Runtime) | 54.3% | 4.1 | 244.3ms | 1537 MB |
Speed Breakdown (A100 TensorRT)
End-to-end latency breakdown showing preprocessing, inference, and postprocessing times
1.1ms
9.6ms
0.1ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import YOLO
# Load model
model = YOLO.from_pretrained("https://huggingface.co/Libre-YOLO/yolov10x")
# Run inference
results = model.predict("image.jpg")
# Process results
for box in results.boxes:
print(f"Class: {box.cls}, Confidence: {box.conf:.2f}")nms-freehighest-accuracy
Notes
Best accuracy-speed tradeoff due to NMS-free design