Ultralytics Platform

Ultralytics Platform is the hosted workspace from the makers of the YOLO object-detection models. It puts the whole vision workflow — upload, annotate, train, export, deploy — in one browser app, so a team can go from a folder of images to a live detection endpoint without setting up GPUs or serving infrastructure.

Data and annotation

• Upload images (up to 50 MB), video (up to 1 GB) or dataset archives from your machine or a URL; Pro and Enterprise also connect Google Cloud Storage, Amazon S3 and Azure buckets.
• The annotation editor covers the six YOLO task types: detection, segmentation, semantic segmentation, oriented boxes, pose and classification.
• Smart Annotation labels for you with SAM models, your own YOLO models, or class-prompted vision-language models from several providers; Ultralytics publishes a leaderboard comparing those models on auto-annotation accuracy and cost.

Training and deployment

• Train YOLO26, YOLO11, YOLOv8 and YOLOv5 on 24 cloud GPU types (26 on paid plans, including B200 and B300), billed per hour with live loss and mAP charts.
• Export to 23 formats, including ONNX, TensorRT, CoreML, LiteRT and Hailo, for edge devices.
• One-click dedicated inference endpoints in 42 regions scale to zero by default and report requests, latency and errors.
• Agents workflows connect models, conditions, dataset collection, Slack alerts and webhooks.

Pricing

Free costs nothing per month and comes with $25 of one-time credits and 100 GB of storage. Pro is $29 per seat a month, includes $30 per seat of monthly credits, 500 GB storage and team collaboration. Enterprise is custom, with SSO, on-premise options and a commercial licence. Free and Pro work under the AGPL-3.0 licence, so shipping YOLO in a closed-source product needs a separate Ultralytics licence.