Ultralytics YOLOv5 models
Build reliable real-time object detection with current YOLOv5u models and the workflow that made training and deploying custom detectors practical and accessible.
YOLOv5u detection models
The current Ultralytics package ships updated anchor-free YOLOv5u detection models in five sizes from nano to extra-large. They keep the proven YOLOv5 workflow while improving the accuracy-speed balance, and they train, validate, export, and deploy through the same API as every Ultralytics YOLO model.
pip install ultralyticsfrom ultralytics import YOLO
# Load a YOLOv5u model
model = YOLO("yolov5nu.pt")
# Run inference
results = model("image.jpg")| Model | Size (px) | mAP 50-95 | CPU ONNX (ms) | A100 TensorRT (ms) | Params (M) | FLOPs (B) |
|---|---|---|---|---|---|---|
| yolov5nu | 640 | 34.3 | 73.6 | 1.06 | 2.6 | 7.7 |
| yolov5su | 640 | 43.0 | 120.7 | 1.27 | 9.1 | 24.0 |
| yolov5mu | 640 | 49.0 | 233.9 | 1.86 | 25.1 | 64.2 |
| yolov5lu | 640 | 52.2 | 408.4 | 2.50 | 53.2 | 135.0 |
| yolov5xu | 640 | 53.2 | 763.2 | 3.81 | 97.2 | 246.4 |
YOLOv5u 640-pixel detection benchmarks from the Ultralytics docs; larger 1280-pixel P6 variants are also available.
Trusted by the world's leading organizations
Our models' impact
Streamline processes across industries with our cutting-edge vision AI models. Speed, accuracy and ease-of-use powered by Ultralytics.
From YOLOv5 to the latest Ultralytics YOLO
See how the practical YOLOv5 workflow grew into a unified model ecosystem for training and deployment.
Made real-time object detection accessible with a fast, practical PyTorch workflow.
Expanded the unified workflow across detection, segmentation, classification, pose, and OBB.
Improved accuracy, speed, and efficiency while preserving the familiar Ultralytics workflow.
Introduced end-to-end inference and an architecture optimized for efficient edge deployment.
Label up to 10x faster with smart annotation
Ultralytics Platform gives you the image annotation tool to build high-quality datasets faster. From smart annotation to precise manual editing, these features are designed to reduce image labeling time without sacrificing quality.
- SAM-powered smart annotation: Masks and bounding boxes in one click.
- Full AI task coverage: Detection, instance segmentation, semantic segmentation, classification, pose, OBB.
- Universal format support: Your choice of YOLO, COCO, VOC, and more.
- Team review and versioning: Clear collaboration at every step.
Train on the Best GPUs for Less
26 NVIDIA GPUs starting at $0.24/hr — from Ampere to Blackwell. No markup, no minimums, no commitment.
Explore industry solutions
See how teams apply Ultralytics computer vision across production environments.
AI in agriculture
AI in automotive
AI in healthcare
AI in logistics
AI in manufacturing
AI in retail
AI in robotics
AI in agriculture
AI in automotive
AI in healthcare
AI in logistics
AI in manufacturing
AI in retail
AI in robotics
AI in agriculture
AI in automotive
AI in healthcare
AI in logistics
AI in manufacturing
AI in retail
AI in robotics
Frequently asked questions
Ultralytics YOLOv5 launched in 2020 with an accessible PyTorch workflow that helped make custom real-time object detection practical. The current Ultralytics Python package supports the updated anchor-free YOLOv5u detection models; original YOLOv5 repository weights use a different architecture and are not compatible with the current library. See the YOLOv5 documentation for details.
The YOLOv5u models supported by the current Ultralytics library and Platform project are built for object detection. The historical YOLOv5 repository also released separate classification and segmentation variants, but those original architectures and weights are not interchangeable with the current YOLOv5u models.
Start with the nano or small YOLOv5u detector when latency and model size matter most. Medium balances speed and accuracy for many workloads, while large and extra-large prioritize accuracy when more compute is available. Compare the current model benchmarks in the YOLOv5 docs, then validate candidates on your own data.
Use the Ultralytics library's train mode with a current YOLOv5u detection checkpoint, or open the official YOLOv5 project on Ultralytics Platform to start from one of its detection models. After training, use export mode to prepare the detector for deployment.
Yes. YOLOv5 remains supported for existing projects and deployments. Ultralytics YOLO26 is the latest stable model and the recommended starting point for new projects, while YOLOv5 remains useful when compatibility with an established YOLOv5 pipeline matters.
Build with Ultralytics YOLOv5
Train, validate, export, and deploy current YOLOv5u detection models with Ultralytics Platform.