Introducing YOLO26

YOLO26 is a multi-task model family supporting object detection, instance segmentation, image classification, pose estimation, and oriented object detection (OBB). It offers models in various sizes from Nano(N) to Extra Large(X) to accommodate diverse deployment environments.

The key improvements are as follows:

• Edge device optimization: Enhanced hardware compatibility through faster CPU inference, compact model design, and a simplified architecture.
• Reduced latency: Removed NMS (Non-Maximum Suppression) to increase inference speed, while maintaining consistent accuracy across fp16 and fp32 environments.
• Efficient architecture: Designed to perform high-performance computer vision tasks with lower latency compared to previous YOLO generations.