Back to Leaderboard

YOLOv8-X

yolov8

one-stage detector with CSPDarknet backbone

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
HardwaremAP@50-95FPSLatencyVRAM
NVIDIA A100 (TensorRT FP16)53.8%63.315.8ms3019 MB
NVIDIA T4 (TensorRT FP16)53.8%23.442.7ms3035 MB
CPU (ONNX Runtime)53.9%2.7369.6ms3018 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

Related Models (yolov8)