How do I deploy Ultralytics YOLO26 on edge devices?

YOLO26 supports export to TensorRT, ONNX, CoreML, TFLite, and OpenVINO, covering the most common edge deployment targets. The NMS-free architecture means fewer integration headaches and lower latency out of the box.

Is Ultralytics YOLO26 compatible with my existing YOLO workflow?

Yes. YOLO26 follows the same interface as previous Ultralytics YOLO releases, so migrating is straightforward. Simply swap in your YOLO26 model weights.

What tasks does Ultralytics YOLO26 support?

Object detection, instance segmentation, semantic segmentation, depth estimation, image classification, pose estimation, and oriented object detection, all in a single unified model family.

Which Ultralytics YOLO26 model size should I use?

The nano (n) variant is ideal for edge and CPU-constrained devices. The small (s) and medium (m) variants offer a strong balance of speed and accuracy for most applications. The large (l) and extra-large (x) variants deliver maximum accuracy for demanding workloads.

What are the key improvements in Ultralytics YOLO26 compared to YOLO11?

YOLO26 removes DFL for simpler export, eliminates NMS for faster end-to-end inference, improves small-object accuracy with ProgLoss + STAL, introduces the MuSGD optimizer for more stable training, and delivers up to 43% faster CPU inference.

Ultralytics YOLO

★★★★★5/5

Ultralytics YOLO26 models

Built from the ground up for edge and low-power devices, Ultralytics YOLO26 sets a new standard for real-time vision AI, delivering up to 43% faster CPU inference with a cleaner, simpler architecture.

Get started

Request an Ultralytics Enterprise License

Trusted by industry leaders

Try YOLO26 inference

Drag and drop an image to see real-time object detection

Input

Upload an image or select an example to test YOLO26n. Predict Docs

Live Demo

Drop an image here

or click to browse

Supports JPEG, PNG, WebP, AVIF, HEIC, JP2, TIFF, BMP & more (max 10MB)

Or try an example

Confidence

0.25

IoU

0.70

Image Size

640

Prediction results will appear here.

Deploy your model to get a dedicated inference endpoint with API key authentication — then replace the placeholder URL and key below with your deployment values. Deploy docs

import requests

# Deployment URL and API key
url = "https://your-deployment-url.run.app/predict"
api_key = "YOUR_API_KEY"

# Optional inference parameters (conf, iou, imgsz)
args = {"conf": 0.25, "iou": 0.7, "imgsz": 640}

with open("image.jpg", "rb") as f:
    response = requests.post(
        url,
        headers={"Authorization": f"Bearer {api_key}"},
        data=args,
        files={"file": f},
    )

print(response.json())

Try all tasks in your browser

Key Ultralytics YOLO26 improvements

Up to 43% faster CPU inference
Real-time performance on devices without GPUs, purpose-built for edge and constrained environments.

NMS-free, end-to-end inference
Predictions are generated directly, with no post-processing step. Lower latency, simpler deployment.

No DFL, broader hardware compatibility
Removing Distribution Focal Loss simplifies exports and broadens edge device compatibility.

MuSGD: A smarter optimizer
A hybrid of SGD and Muon inspired by LLM training advances for more stable training.

Ultralytics YOLO models at a glance

Compare recent Ultralytics YOLO model families across speed, accuracy, and supported vision tasks.

Features
Speed Image processing time	0.54 ms	0.47 ms	0.56 ms	0.67 ms	0.62 ms
Accuracy mAP_50-95	34.3%	37.3%	39.5%	40.9%	42.3%
Object detection
Instance segmentation	-
Semantic segmentation	-		-
Depth estimation	-	-	-
Image classification	-
Pose estimation	-
OBB detection	-

License YOLO

Security & compliance

Enterprise-grade security and compliance

Build and deploy with confidence. ISO 27001 certified, SOC 2 attested, with support for HIPAA compliance.

License YOLO Explore security

SOC 2 Type I

Attested against AICPA Security, Availability, and Confidentiality TSC

ISO 27001:2022

Certified information security management system

HIPAA

Security Rule self-assessment; BAAs for eligible Enterprise customers

GDPR & CCPA

Aligned with global data protection regulations

Why choose Ultralytics YOLO26?

Built for edge and cloud
Runs efficiently on CPUs, GPUs, and edge hardware. Export to 22 formats and deploy anywhere.

Up to 43% faster CPU inference
Real-time vision AI on resource-constrained devices, without sacrificing accuracy.

Open-world detection with YOLOE-26
Detect beyond fixed categories using text prompts, visual prompts, or prompt-free inference.

Seamless integration
YOLO26 follows the same familiar interface as YOLOv8 and YOLO11, with no steep learning curve.

Backed by a global community
Dedicated support channels, active forums, and regular updates keep you moving forward.

Clear licensing
Flexible options for academic, open-source, and commercial use under AGPL-3.0 and Enterprise.

What our clients say

Ultralytics YOLO models helped us boost our AI service implementation speed by 25.2%. They are easy to deploy and provide excellent real-world performance, delivering greater speed, accuracy, and scalability without compromising operational performance.

Of all the models we evaluated, YOLO was by far the fastest for our use cases.

Ultralytics YOLO26 runs seamlessly on Axelera's platform, enabling faster edge AI deployment with minimal integration effort. This is exactly the kind of hardware-software collaboration that helps customers move from development to production faster.

Alexis Crowell, Global CMO & GM of the Americas

The ability to train and deploy Ultralytics YOLO models through the Ultralytics Python package is very solid. Combined with fast inference, it enables us to process images quickly and deliver real-time feedback on construction sites.

We make computer vision-enabled robots to pick and place ingredients on food production lines. Robotic manipulation demands below-centimeter precision, and Ultralytics YOLO models are great at providing that level of accuracy.

The performance of Ultralytics YOLO models on our STM32 microcontrollers opens up new horizons for computer vision at the edge. With streamlined integration and support for multiple model variants, developers have exactly what they need to build scalable and responsive embedded vision solutions.

The Ultralytics Python package allows us to train models very easily and quickly. Furthermore, its standardization across different models and new versions of YOLO ensures a high degree of stability over the years.

Ultralytics YOLO26 models

Try YOLO26 inference

Key Ultralytics YOLO26 improvements

Up to 43% faster CPU inference

NMS-free, end-to-end inference

No DFL, broader hardware compatibility

MuSGD: A smarter optimizer

Ultralytics YOLO models at a glance

Enterprise-grade security and compliance

SOC 2 Type I

ISO 27001:2022

HIPAA

GDPR & CCPA

Why choose Ultralytics YOLO26?

Built for edge and cloud

Up to 43% faster CPU inference

Open-world detection with YOLOE-26

Seamless integration

Backed by a global community

Clear licensing

What our clients say

See our models in action

RTM cuts inference cost 28% with Ultralytics YOLO26

SK Godelius detects 3,400 defects on mining conveyors with Ultralytics YOLO

Camerite cuts development time 25.2% with Ultralytics YOLO

OnsiteIQ reduces project delays by 20% with Ultralytics YOLO11

mtrail's TrainVision cuts rail data collection time by close to 100%

Glacier Robotics reduced PET leakage by 70% across US recycling facilities

SOHGA cuts parking monitoring time by 30% with Ultralytics YOLO

Scaleout cuts model updates from weeks to hours with Ultralytics YOLO

RapiD Engineering deploys seafood quality control 1 week faster with Ultralytics YOLO

Project Ocean Oasis advances reef conservation with Ultralytics YOLO

Volley powers 250+ on-court AI trainers with Ultralytics YOLO

WG Tech Solutions cuts safety violations by 28% with Ultralytics YOLO and Axelera’s AI Accelerator

Stride delivers 1-minute equine gait analysis with Ultralytics YOLO

Pixelabs achieves 95% recall with Ultralytics YOLO-driven automation

SiteAssist improves site safety by processing 770K+ images with Ultralytics YOLO

Chef Robotics uses Ultralytics YOLO to cut food giveaway by 67%

Cali Intelligence shortens checkout queues by 43% with Ultralytics YOLO

MarineSitu hits 96%+ uptime in underwater monitoring using Ultralytics YOLO

Theia Scientific speeds up microscopy analysis 43× with Ultralytics YOLO

eSmart Systems halves power line inspection time with Ultralytics YOLO

Axelera AI delivers 34 FPS edge AI inference using Ultralytics YOLO

STMicroelectronics runs Ultralytics YOLO on an MCU at just 9.4 mJ per inference

Specialvideo reaches 99% food inspection accuracy with Ultralytics YOLO

Vivity AI saves $5M+ a year in industrial operations with Ultralytics YOLO

Videologic Analytics scales to 10K AI camera licenses with Ultralytics YOLO

Prezent boosts slide detection accuracy by 34% with Ultralytics YOLO

ALYCE accelerates traffic AI inference by 20% with Ultralytics YOLO

Kiwitron uses Ultralytics YOLO to detect industrial hazards 30m away

Frequently asked questions

How do I deploy Ultralytics YOLO26 on edge devices?

Is Ultralytics YOLO26 compatible with my existing YOLO workflow?

What tasks does Ultralytics YOLO26 support?

Which Ultralytics YOLO26 model size should I use?

What are the key improvements in Ultralytics YOLO26 compared to YOLO11?

Get started with Ultralytics YOLO!