Ultralytics support
Find answers in our documentation, connect with the community, or reach out to our team directly.
Get help by topic
Find your problem area below for step-by-step guides, tutorials, and troubleshooting.
Annotate
Labeling, dataset formats, and annotation troubleshooting
Learn more
Train
Training setup, GPU and CUDA errors, and improving accuracy
Learn more
Deploy
Export formats, edge and cloud deployment, and inference
Learn more
Agents
Visual workflows with YOLO, vision-language models, conditions, and alerts
Learn more
Ultralytics YOLO models
Choosing a model, prediction modes, and architecture guides
Learn more
API
SDK setup, REST API reference, and integration help
Learn more
Ultralytics YOLO license
AGPL-3.0 vs Enterprise, compliance, and billing
Learn moreFrom the blog
Gemini 4 Carbon auto-annotation: the new leader on our VLM Leaderboard
Gemini 4 Carbon tops 45 VLMs on our 36-dataset object detection benchmark: 0.584 mAP50 and the best precision and recall. When GPT-6.1 Terra is better value.
How to Turn YOLO Detections into Alerts Without Code Using Ultralytics Agents
Turn Ultralytics YOLO detections into Slack alerts, webhooks and dataset captures without code, using a Gate condition in Ultralytics Agents.
Budgeting for Image Annotation: Seats, AI and Exports
Compare image annotation software costs, free-plan limits and AI allowances. Estimate software and project costs with a worked three-person team example.
Claude Fable 5.2 is here, and it refuses to look at security X-rays
Claude Fable 5.2 is out, and it refuses to label the IEDXray security X-ray dataset on Ultralytics Platform: all 10 scans blocked, while open models answered.
Small object detection with image tiling in Ultralytics YOLO, from Python to Rust
See when image tiling helps YOLO detect small objects, compare its costs, and run built-in tiled inference in Python or Rust without SAHI.
Best platforms for training custom image classifiers
Image classifier training platforms compared: no-code AutoML, developer platforms and edge tools, and how to choose by control, deployment and cost.
How to tune the NMS IoU threshold in Ultralytics YOLO
Tune the NMS IoU threshold in Ultralytics YOLO: why prediction now defaults to 0.5, when to change iou vs. conf, and when to skip NMS with YOLO26.
Best synthetic data tools for computer vision training
Synthetic data tools for computer vision compared by method: 3D simulation, generative models and augmentation, and how to test them with Ultralytics YOLO.
Ultralytics at AMD Embedded AI Working Group in Dublin
Discover how Ultralytics and AMD explored Physical AI, robotics, and computer vision deployment at AMD Embedded AI Working Group 2026 in Dublin.
How to choose a VLM for auto-annotation using precision and recall
Pick a VLM for auto-annotation by precision and recall: which models leave fewer stray boxes, which miss fewer objects, and how to test on your own data.
ONNX vs TensorRT for computer vision inference: benchmarks and how to choose
ONNX Runtime vs TensorRT for computer vision inference: measured Ultralytics YOLO26 speed and accuracy on 19 NVIDIA GPUs, and how to choose.
Missing PPE detection: train Ultralytics YOLO on Construction-PPE
See how Construction-PPE labels missing helmets, vests and gloves, then train Ultralytics YOLO to detect missing PPE on your own site cameras.
Drone detection dataset guide: train Ultralytics YOLO on tiny UAVs
Compare DUT Anti-UAV, Drone-vs-Bird and thermal drone detection datasets, then train Ultralytics YOLO26 to find tiny drones and cut false alarms.
HIPAA compliant AI for medical imaging: what a BAA covers and how to train on patient data
Learn what HIPAA compliant AI requires for medical imaging, what an Ultralytics BAA covers, and how to annotate, train and deploy with patient data.
Best Depth Estimation Models: How to Choose
Compare monocular depth models such as Depth Anything, Depth Pro, Metric3D, UniDepth and Ultralytics YOLO26 for metric depth, edge devices and robotics.
How to migrate Ultralytics YOLO models from W&B to Ultralytics Platform
Import Ultralytics YOLO models from Weights & Biases (W&B) into Ultralytics Platform without retraining, then compare experiments, export, and deploy.
GPT-6.1 Terra takes #1 for auto-annotation, beating OpenAI's own flagships
GPT-6.1 Terra auto-annotation tops 39 models on our VLM Leaderboard: 0.560 mAP50 for $4.36 per 1,000 images, ahead of GPT-6.1 Sol and Gemini 4 Argon.
MiniMax-M3.1-Flash auto-annotation: why less reasoning drew better boxes
MiniMax-M3.1-Flash auto-annotation in Ultralytics Platform: 0.317 mAP50 at $1.84 per 1,000 images after tuning, with 0.86 on SAR ship detection.
Kimi K3.1 Pro auto-annotation: a third better than K3, at high effort
Kimi K3.1 Pro auto-annotation scores 0.444 mAP50 on UL37, a third above Kimi K3. See why we run it at high effort, what it costs, and where it wins.
Ultralytics YOLO and NVIDIA Cosmos Data Factory: Synthetic data for defect detection
NVIDIA Cosmos Data Factory synthetic data and an agentic training loop help Ultralytics YOLO26 match full-dataset defect detection accuracy with fewer labels.
Gemini 4 Argon auto-annotation: the most precise boxes of 39 models
Gemini 4 Argon auto-annotation on Ultralytics Platform: #2 of 39 models, top precision (0.667), first in 6 of 12 domains, $7.60 per 1,000 images to Nov 30.
Open Compute, Open Vision: AMD at Ultralytics YOLO Vision 2026
See how AMD brought open compute to Ultralytics YOLO Vision 2026 with a ROCm talk, hands-on workshops, and demos running YOLO on AMD Radeon GPUs.
Best Image Classification Models: How to Choose
Compare the leading image classification models, from ConvNeXt V2 and DINOv3 to EfficientNetV2, ViT and Ultralytics YOLO26-cls, and learn how to choose.
Semantic vs Instance Segmentation: How to Choose the Right Task
Semantic segmentation labels every pixel by class; instance segmentation separates each object. Learn which one your computer vision project needs, and why.
Ultralytics at Arm Create 2026: Real-time Vision AI at the edge
Highlights from Ultralytics at Arm Create 2026 in Shenzhen: deploying Ultralytics YOLO for real-time edge AI, the model panel, and the new Arm AI Portal.
Ultralytics Meetup: London
Recap of the Ultralytics London meetup: A first look at YOLO27, Ultralytics Platform demos, and a vision AI panel with NVIDIA and Arm.
Ultralytics Platform Monitoring: Keep vision AI deployments on track
Monitor deployed vision AI models with Ultralytics Platform. Track endpoint health, latency, errors, logs, and live prediction statistics.
Introducing Ultralytics Agents: Build visual AI workflows faster
Build visual AI workflows with Ultralytics Agents. Connect YOLO models, vision-language models, conditions, datasets, and Slack alerts on one canvas.
Introducing Ultralytics AutoTrain: Simplify vision AI model training
Train, compare, and improve Ultralytics YOLO models with AutoTrain in Ultralytics Platform, from initial baseline to production deployment.
Key highlights from Ultralytics YOLO Vision 2026!
Key moments from YOLO Vision 2026, where Ultralytics unveiled Ultralytics YOLO27 and showcased AutoTrain, Agents, and Monitoring on Ultralytics Platform.
Frequently asked questions
Yes. Beyond standard support, Ultralytics offers custom engineering services for teams that need hands-on help integrating YOLO into web, mobile, or cloud products. This includes model training, deployment architecture, and bespoke solutions. Contact the sales team to discuss your requirements and get a quote.
For bugs with the open-source YOLO package, open an issue on GitHub. For issues specific to Ultralytics Platform, such as training jobs, deployments, or billing, use the feedback form inside the platform or reach out via your dedicated support channel if you're on an Enterprise plan.
For community support, response times vary based on activity in the forum and GitHub, though our team aims to respond to GitHub issues within 24 hours. Enterprise customers receive guaranteed response times defined in a contractual SLA, tailored to the size and needs of their organization.
Enterprise customers receive a named SE/CSM, contractual SLA, SE enablement hours, structured onboarding with milestone checkpoints, project enablement support, and quarterly business reviews. It's designed for organizations running Ultralytics at scale who need a direct line to our team from day one. Contact sales to discuss what's right for your organization.
Free and Pro plans include access to community support, our active forum where you can ask questions, share projects, and get answers from other developers and the Ultralytics team. You also have full access to Ultralytics documentation, GitHub, and video tutorials covering everything from annotation to deployment.
Let's build the future of AI together!
Begin your journey with the future of machine learning