Back to Leaderboard

YOLOv8-L

yolov8

one-stage detector with CSPDarknet backbone

Parameters

43.7M

FLOPs

165.2G

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)52.9%82.512.1ms1862 MB
NVIDIA T4 (TensorRT FP16)53.0%32.031.2ms1883 MB
CPU (ONNX Runtime)53.0%3.7272.0ms1911 MB
Speed Breakdown (A100 TensorRT)
End-to-end latency breakdown showing preprocessing, inference, and postprocessing times
1.4ms
8.1ms
2.7ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import YOLO

# Load model
model = YOLO.from_pretrained("https://huggingface.co/Libre-YOLO/yolov8l")

# Run inference
results = model.predict("image.jpg")

# Process results
for box in results.boxes:
    print(f"Class: {box.cls}, Confidence: {box.conf:.2f}")
production-readyhigh-accuracy
Related Models (yolov8)