Try the Model

Use the widget below to experiment with YOLO26. You can detect COCO classes such as people, vehicles, animals, household items.

Overview

YOLO26 is a real-time computer vision model released in late October 2025. It introduces a native end-to-end, NMS-free architecture that removes the need for a separate Non-Maximum Suppression post-processing step, simplifying inference pipelines and reducing latency, particularly on edge and low-power hardware.

YOLO26 incorporates several training and architectural updates, including the MuSGD optimizer and the use of ProgLoss with STAL to improve training stability and small-object performance. The model removes Distribution Focal Loss (DFL), improving compatibility with export targets such as ONNX and TensorRT. YOLO26 supports multiple vision tasks—object detection, segmentation, pose estimation, and oriented bounding box detection—within a unified framework and is available in model sizes ranging from Nano to Extra Large.

‍

YOLO26 License

YOLO26

is licensed under a

AGPL-3.0

license.

Performance

Deploy a YOLO26 API

You can use Roboflow Inference to deploy a

YOLO26

API on your hardware. You can deploy the model on CPU (i.e. Raspberry Pi, AI PCs) and GPU devices (i.e. NVIDIA Jetson, NVIDIA T4).

Below are instructions on how to deploy your own model API.

Label Data Automatically with YOLO26

You can automatically label a dataset using YOLO26 with help from Autodistill, an open source package for training computer vision models. You can label a folder of images automatically with only a few lines of code. Below, see our tutorials that demonstrate how to use YOLO26 to train a computer vision model.

No items found.