Two places to run a model
Cloud deployment sends data to servers you rent, where models can be large and easy to update. Edge deployment runs the model on or near the device that captures the data — a phone, a camera, a small computer on site.
What pushes toward the edge
Run at the edge when:
- Decisions need to happen in milliseconds.
- Connectivity is poor or expensive.
- Raw video or images should not leave the site.
- Bandwidth to stream everything would cost too much.
| Edge | Cloud | |
|---|---|---|
| Latency | Lowest | Network-dependent |
| Model size | Constrained by the device | Large models available |
| Updates | Fleet rollout | Deploy once |
| Data movement | Stays local | Sent to the provider |
What pushes toward the cloud
Run in the cloud when the model is large, when you need to update it often, when many sources feed one system, or when a person reviews results centrally. Many systems do both: a small model filters at the edge and sends only the interesting moments to the cloud.