Cloud Model

What is a Cloud Model?

A Cloud Model runs on a provider’s infrastructure. Requests are sent through an API to the provider’s servers, processed there, and the response is returned. Unlike a Local Model, the user does not have to operate or maintain the underlying infrastructure, but in return hands over their requests to the provider.

How it works

A request is sent over the Internet to a provider’s API. Inference is performed there on specialized AI hardware (often GPU clusters), which goes far beyond what typical end-user devices can provide. The response is returned over the same connection, with billing usually based on the number of tokens processed.

Why choosing between cloud and local is a trade-off

Cloud Models provide access to significantly larger and more powerful models than could typically be operated locally, without requiring investment in dedicated hardware. The trade-off is that requests leave the user’s own infrastructure, an Internet connection is required, and usage is subject to the provider’s terms and pricing.

A Cloud Model is not automatically less secure than a Local Model. Many providers implement security measures that are far more extensive than what an individual user could achieve locally. The more relevant question is whose control over the data is preferred—your own infrastructure or the provider’s infrastructure—not which option is technically more secure in general.

Why Cloud Models have practical implications

• Performance: Access to model sizes that would not normally run on typical end-user hardware.
• Dependency: Use requires an Internet connection and depends on the provider’s availability.
• Cost: Usually billed per token instead of requiring a one-time hardware investment.

Typical example

A request to a large model that cannot realistically be run locally is sent through an API, processed on the provider’s servers, and returned within a few seconds, regardless of the user’s own hardware.

Further reading

Fundamentals

→ Local AI Models: Concepts and Requirements

Related terms

→ Local Model
→ Inference