Meet YOLO26: the latest stable Ultralytics YOLO 🚀

[glenn-jocher]

YOLO26 is the current recommended Ultralytics model generation for real-time vision AI: faster, more accurate, production-ready, and available across the core YOLO task family.

This pinned issue is the launch hub for users who want the latest Ultralytics models, examples, docs, and upgrade path in one place.

What shipped

• YOLO26 detection with end-to-end, NMS-free inference support for simpler deployment paths.
• Six vision tasks in one framework: detection, instance segmentation, semantic segmentation, classification, pose estimation, and oriented bounding boxes.
• Five model scales: n, s, m, l, and x, from edge devices to high-accuracy workloads.
• Automatic weight downloads from official Ultralytics releases on first use.
• Production workflows for train, val, predict, track, benchmark, and export.

Start here

• YOLO26 docs: https://docs.ultralytics.com/models/yolo26
• Roadmap: https://www.ultralytics.com/roadmap
• Main repository: https://github.com/ultralytics/ultralytics
• Python usage: https://docs.ultralytics.com/usage/python
• CLI usage: https://docs.ultralytics.com/usage/cli
• Configuration: https://docs.ultralytics.com/usage/cfg
• Train: https://docs.ultralytics.com/modes/train
• Predict: https://docs.ultralytics.com/modes/predict
• Track: https://docs.ultralytics.com/modes/track
• Export: https://docs.ultralytics.com/modes/export
• Ultralytics Platform: https://platform.ultralytics.com/ultralytics/yolo26

Quickstart: CLI

pip install -U ultralytics

# Predict with the smallest YOLO26 detection model
yolo predict model=yolo26n.pt source=https://ultralytics.com/images/bus.jpg imgsz=640

# Train on a small detection dataset
yolo train data=coco8.yaml model=yolo26n.pt epochs=10 imgsz=640

# Validate a pretrained model
yolo val data=coco8.yaml model=yolo26n.pt imgsz=640

# Export for deployment
yolo export model=yolo26n.pt format=onnx imgsz=640

Quickstart: Python

from ultralytics import YOLO

# Load an official YOLO26 model. Weights download automatically on first use.
model = YOLO("yolo26n.pt")

# Run inference
results = model("https://ultralytics.com/images/bus.jpg", imgsz=640)

# Train, validate, and export
model.train(data="coco8.yaml", epochs=10, imgsz=640)
metrics = model.val(data="coco8.yaml", imgsz=640)
model.export(format="onnx", imgsz=640)

Task examples

# Detection
yolo predict model=yolo26n.pt source=https://ultralytics.com/images/bus.jpg

# Instance segmentation
yolo segment predict model=yolo26n-seg.pt source=https://ultralytics.com/images/bus.jpg

# Semantic segmentation
yolo semantic predict model=yolo26n-sem.pt source=https://ultralytics.com/images/bus.jpg imgsz=1024

# Classification
yolo classify predict model=yolo26n-cls.pt source=https://ultralytics.com/images/bus.jpg

# Pose estimation
yolo pose predict model=yolo26n-pose.pt source=https://ultralytics.com/images/bus.jpg

# Oriented bounding boxes
yolo obb predict model=yolo26n-obb.pt source=https://ultralytics.com/images/boats.jpg

# Multi-object tracking
yolo track model=yolo26n.pt source=https://youtu.be/LNwODJXcvt4 tracker=botsort.yaml

Useful model and dataset links

• Detection task: https://docs.ultralytics.com/tasks/detect
• Instance segmentation task: https://docs.ultralytics.com/tasks/segment
• Semantic segmentation task: https://docs.ultralytics.com/tasks/semantic
• Classification task: https://docs.ultralytics.com/tasks/classify
• Pose task: https://docs.ultralytics.com/tasks/pose
• OBB task: https://docs.ultralytics.com/tasks/obb
• COCO8 quick test dataset: https://docs.ultralytics.com/datasets/detect/coco8
• Cityscapes semantic dataset: https://docs.ultralytics.com/datasets/semantic/cityscapes
• Model comparison pages: https://docs.ultralytics.com/compare/yolo26-vs-yolo11

Need help?

For usage questions, include your command, model name, dataset YAML, and yolo checks output. For bugs, please open a focused issue with a minimum reproducible example: https://docs.ultralytics.com/help/minimum-reproducible-example

YOLO26 is the recommended starting point for new Ultralytics projects. Older YOLOv3 and YOLOv5 repositories remain available for legacy workflows, but new models and active development live here in ultralytics/ultralytics.

[UltralyticsAssistant]

👋 Hello @glenn-jocher, thank you for opening this YOLO26 documentation tracking issue 🚀 This is an automated response to help with triage, and an Ultralytics engineer will also assist soon.

We recommend a visit to the Docs for new users where you can find many Python and CLI usage examples and where many of the most common questions may already be answered.

For this documentation-focused issue, please continue to add clear notes on any gaps around model selection, pretrained weights, training arguments, export behavior, deployment paths, and reproducibility details 📝

If this is a 🐛 Bug Report, please provide a minimum reproducible example to help us debug it. For code bugs, opening a separate focused report with an MRE will help us triage it efficiently.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset image examples and training logs, and verify you are following our Tips for Best Training Results.

Join the Ultralytics community where it suits you best. For real-time chat, head to Discord 🎧. Prefer in-depth discussions? Check out Discourse. Or dive into threads on our Subreddit to share knowledge with the community.

Upgrade

Upgrade to the latest ultralytics package including all requirements in a Python>=3.8 environment with PyTorch>=1.8 to verify your issue is not already resolved in the latest version:

pip install -U ultralytics

Environments

YOLO may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all Ultralytics CI tests are currently passing. CI tests verify correct operation of all YOLO Modes and Tasks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

[ldenoue]

is YOLO26 exportable to Apple's new Core AI?

I saw their repo but unfortunately only 2 models are currently supported:
https://github.com/apple/coreai-models/tree/main/models/yolo

Laurent

[glenn-jocher]

Yes for Apple Core ML today: yolo export model=yolo26n.pt format=coreml quantize=8 produces a .mlpackage that can run on Apple devices, and the Ultralytics CoreML guide covers the recommended iOS/Flutter deployment path. Apple Core AI .aimodel export is a separate Apple toolchain, and it is not currently a built-in Ultralytics export target, so for YOLO26 on Apple hardware CoreML is the supported path right now.