Deploy computer vision models across 43 global regions

Yes. Each model can be deployed to multiple regions simultaneously. Your plan determines the total number of endpoints available: 3 for Free, 10 for Pro, and unlimited for Enterprise. This allows you to serve users globally with low-latency endpoints in each region.

Dedicated endpoints are billed based on CPU, memory, and request volume. With scale-to-zero enabled by default, you only pay for active inference time. There is no cost when your endpoint isn't receiving requests. Shared inference is included with your platform plan.

Shared inference runs on a multi-tenant service across 3 regions and is rate-limited to 20 requests per minute. It's best for development and quick testing. Dedicated endpoints are single-tenant services deployed to any of 43 regions with no rate limits, consistent latency, and configurable resources, built for scalable production workloads.

Dedicated endpoint deployment typically takes one to two minutes. This includes container provisioning, startup, and an initial health check to validate the service is ready. Once the endpoint is ready, it begins accepting inference requests immediately.

Model deployment is the process of making a trained computer vision model available to receive and process real-world data. Once deployed, computer vision applications can send images and video frames to the model via API and receive predictions, enabling everything from automated quality inspection to real-time object detection in production systems. On Ultralytics Platform, deployment is integrated directly into the end-to-end training workflow. Once your model is trained, you can test it in the browser, deploy it to a dedicated endpoint in any of 43 global regions, and monitor its performance, all from the same workspace.