Back to Leaderboard
Parameters
68.2M
FLOPs
257.8G
Input Size
640px
License
MIT
Architecture
Type
one-stage
Backbone
CSPDarknet
Neck
PAFPN
Head
Decoupled
Benchmark Results
Performance on COCO val2017 across different hardware configurations
| Hardware | mAP@50-95 | FPS | Latency | VRAM |
|---|---|---|---|---|
| NVIDIA A100 (TensorRT FP16) | 53.8% | 63.3 | 15.8ms | 3019 MB |
| NVIDIA T4 (TensorRT FP16) | 53.8% | 23.4 | 42.7ms | 3035 MB |
| CPU (ONNX Runtime) | 53.9% | 2.7 | 369.6ms | 3018 MB |
Speed Breakdown (A100 TensorRT)
End-to-end latency breakdown showing preprocessing, inference, and postprocessing times
1.1ms
12.4ms
2.3ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import YOLO
# Load model
model = YOLO.from_pretrained("https://huggingface.co/Libre-YOLO/yolov8x")
# Run inference
results = model.predict("image.jpg")
# Process results
for box in results.boxes:
print(f"Class: {box.cls}, Confidence: {box.conf:.2f}")production-readyhighest-accuracy
Notes
Largest YOLOv8 variant, best accuracy in the family