YOLO26L is the large-sized variant of the YOLO26 model family, designed to deliver high detection accuracy while still supporting real-time inference on capable hardware.
Sitting between the medium and extra-large models, YOLO26L provides a strong balance for applications that demand robust accuracy in complex scenes without fully committing to the computational cost of the largest model. It uses the same end-to-end, NMS-free architecture, enabling a streamlined and consistent deployment experience.
YOLO26L is pretrained on the COCO dataset and is well-suited for production-grade vision systems where accuracy is a key requirement.
Large model size with strong accuracy improvements
End-to-end detection with NMS-free inference
Real-time capable on modern GPUs and high-performance accelerators
Pretrained weights available for immediate use
Ultralytics ecosystem support for training, validation, inference, and export
Use Cases
High-accuracy real-time video analytics
Advanced robotics and automation
Smart city and traffic monitoring
Industrial inspection and safety systems
Production computer vision pipelines
Inputs and Outputs
Input:
Images or video streams, automatically preprocessed by the Ultralytics framework
Output:
Bounding boxes
Class labels
Confidence scores
License
This repo is licensed under the Creative Commons Attribution–NonCommercial 4.0 (CC BY-NC 4.0) license, which allows use, sharing, and modification only for non-commercial purposes with proper attribution. All NPU-related models, runtimes, and code in this project are protected under this non-commercial license and cannot be used in any commercial or revenue-generating applications. Commercial licensing or enterprise usage requires a separate agreement. For inquiries, please contact dev@nexa.ai
