YOLOv8
Ultralytics/YOLOv8
published Jan 2024 · updated Jun 2026
YOLOv8 is a state-of-the-art object detection model that is fast, accurate, and easy to use for a variety of computer vision tasks.
specs
| Task | Object Detection |
| Architecture | YOLOv8 (CSPDarknet backbone, PANet neck) |
| Parameters | 3.2M (nano) to 68.2M (extra-large) |
| License | AGPL-3.0 |
about this model
Benchmark Performance
The following performance metrics are measured on the COCO val2017 dataset at 640-pixel input size. Speed values are averaged over COCO val images using an Amazon EC2 P4d instance.
| Model | mAP | Speed CPU ONNX (ms) | Speed A100 TensorRT (ms) | Params (M) | FLOPs (B) |
|---|---|---|---|---|---|
| YOLOv8n | 37.3 | 80.4 | 0.99 | 3.2 | 8.7 |
| YOLOv8s | 44.9 | 128.4 | 1.20 | 11.2 | 28.6 |
| YOLOv8m | 50.2 | 234.7 | 1.83 | 25.9 | 78.9 |
| YOLOv8l | 52.9 | 375.2 | 2.39 | 43.7 | 165.2 |
| YOLOv8x | 53.9 | 479.1 | 3.53 | 68.2 | 257.8 |
best for
- ·Real-time object detection in video streams
- ·Industrial defect detection
- ·Autonomous driving perception
FAQ
YOLOv8 excels at real-time object detection, instance segmentation, pose estimation, and classification, with a focus on speed and accuracy.
YOLOv8 introduces architectural improvements like a new backbone and neck, offering higher mAP and faster inference than YOLOv5 and YOLOv6, with multiple model sizes to balance speed and accuracy.
YOLOv8 accepts images (e.g., JPEG, PNG) and returns bounding boxes with class labels and confidence scores. The model is trained on 640x640 pixel images.
Send a POST request to the gigarouter OpenAI-compatible endpoint with an API key, providing the image as base64 or URL, and receive detection results in JSON format.
YOLOv8 is released under the AGPL-3.0 license. Commercial use requires an enterprise license from Ultralytics.
We're benchmarking and onboarding YOLOv8 as a hosted, OpenAI-compatible API. Sign in for free credit and be ready when it lands, or tell us you want it and we'll prioritize it.