Skip to content
imaged

Detect, segment, pose, OBB, classify

YOLO26x

The Ultralytics flagship covering five tasks in one family.

AGPL-3.0
Explore model

About this model

YOLO26x is the extra-large variant of the YOLO26 release. The same family auto-detects which task to run from the output shape: detection, segmentation, pose with 17 keypoints, oriented bounding boxes, and 1000-class ImageNet classification.

On M1 Max, the detect ONNX runs in 79 ms warm with CoreML, about 13 fps. The detect ONNX is 223 MB; the other task heads ship as separate weights from the same architecture.

What you can build

  • Run accuracy-first detection where YOLO11n misses small objects.
  • Get instance masks for product photos before background removal.
  • Estimate human pose with 17 COCO keypoints for sports analysis.
  • Detect rotated objects from aerial imagery with the OBB head.
  • Drop in 1000-class ImageNet classification from the same family.

Variants & downloads

Choose the build for your app

Model files are being prepared.

Explore the model’s examples, source links, and integration notes while download variants are added.

Example inference

Input
The canonical bus and pedestrians fixture.
Output
Detection boxes drawn on the bus scene.

Performance

Results depend on the device, runtime, and test conditions. Each measurement keeps its own context.

M1 Max, macOS 15, ONNX Runtime 1.17

See original catalog hardware notes · CPU

641.4 ms

Warm inference

Catalog reference · variant unspecifiedCold start (ms): 658.21

Reference result from the existing catalog; conversion variant and measurement date were not recorded.

M1 Max, macOS 15, ONNX Runtime 1.17

See original catalog hardware notes · CoreML

78.7 ms

Warm inference

Catalog reference · variant unspecifiedCold start (ms): 175.99

Reference result from the existing catalog; conversion variant and measurement date were not recorded.

RTX 2060, Ubuntu 24.04, ONNX Runtime 1.24.3 (detect head)

See original catalog hardware notes · CUDA

37.8 ms

Warm inference

Catalog reference · variant unspecified

Reference result from the existing catalog; conversion variant and measurement date were not recorded.

Integration examples

Vision SDK integration

cpp

#include "imaged/sdk.hpp"
#include "imaged/families/yolo/yolo_family.hpp"

imaged::SDK::AI ai;
auto& config = ai.getOptions();
config.set("yolo.model", "weights/yolo26x.onnx");
config.set("yolo.confidenceThreshold", 0.25);

imaged::SDK::YoloFamily family;
family.load(config);

imaged::SDK::Image image;
image.load("frame.jpg");

auto result = family.run({&image});
auto& detections = std::get<imaged::SDK::DetectionResult>(result).detections;