What it means
Model hosting describes where and how an AI model is run so that applications can use it, including the servers, operating responsibilities and the interface through which requests are sent. Models can be hosted by the model provider, by a cloud platform, by a specialised service provider or by the organisation itself.
Real-world example
A state government wants to use the same language model across several ministries. It compares three hosting options: the model provider's own service, the same model offered through a large cloud platform's EU region, and a dedicated instance run by the state's own IT service provider. Each option differs in cost, data flows, contracts and who is responsible for updates.
Why it matters for AI adoption
The same model can come with very different data protection, security, cost and availability conditions depending on who hosts it. Hosting decisions shape contracts, risk assessments and how easily the organisation could switch models later.
Questions to ask
- Who hosts the model, and who is responsible for security, updates and availability?
- How easily could we move to a different model or host?