YOLO26-S

yolo26Coming Soon

one-stage detector with CSPDarknet backbone

Parameters9.5M

GFLOPs20.7

Input Size640px

Best mAP47.4%

Licensenon-permissive

Architecture

Type

one-stage

Backbone

CSPDarknet

Neck

PAFPN

Head

NMS-Free End-to-End

Benchmark Results

Performance on COCO val2017 across different hardware configurations

Hardware	Runtime	mAP@50-95	FPS	Latency	VRAM
NVIDIA Jetson Orin Nano Super 8GB	ONNX Runtime FP32	47.4%	4.0	247.5ms	-
NVIDIA Jetson Orin Nano Super 8GB	PyTorch FP32	47.4%	20.1	49.9ms	353 MB
NVIDIA Jetson Orin Nano Super 8GB	TensorRT FP16	47.4%	41.7	24.0ms	11 MB
Raspberry Pi 5	ncnn FP32	47.4%	5.7	176.9ms	-
Raspberry Pi 5	ONNX Runtime FP32	47.4%	2.8	360.4ms	-
Raspberry Pi 5	PyTorch FP32	47.4%	2.3	441.8ms	-

Speed Breakdown(NVIDIA Jetson Orin Nano Super 8GB)

7.0ms

233.9ms

0.7ms

Preprocess

Inference

Postprocess (NMS)

anchor-freenms-freePaper: 48.6% mAP

Related Models (yolo26)

YOLO26-N YOLO26-M YOLO26-L YOLO26-X

Run any model with one line

LibreYOLO has the best catalogue of state-of-the-art detectors, all behind one MIT-licensed Python API.

from libreyolo import LibreYOLO, SAMPLE_IMAGE # LibreYOLO has the best catalogue of state-of-the-art models. model = LibreYOLO("LibreRFDETRl.pt") # RF-DETR-L (transformer flagship) results = model(SAMPLE_IMAGE, save=True) # run inference, save the annotated image # Swap in any other model, same one-line API (weights auto-download): # LibreYOLO("LibreYOLO9c.pt") # YOLO9-C # LibreYOLO("LibreYOLOXx.pt") # YOLOX-X # LibreYOLO("LibreDFINEx.pt") # D-FINE-X # LibreYOLO("LibreRTDETRr50.pt") # RT-DETR-R50