Generally, a 'model' is a description of something. A model of a car, or a model of a house, etcetera. A maquette, being a model of a house, can be the starting point for actually creating the building in brick and mortar. But in Perspectives, a Model is *executable*.
A Perspectives model contains all essential information to construct a real, working application. But we need not construct that application by hand. Perspect IT has built an *interpreter* that reads a model and automatically constructs screens for the user to interact with.

The Perspectives Distributed Runtime (PDR)

This interpreter is part of the Perspectives Distributed Runtime (PDR). This is a program that runs on your computer, tablet or smartphone. It connects with other users, that is, their respective instances of the PDR. These PDR's automatically exchange information according to the rules laid down in the model. So if a user updates some property, the PDR decides, on the basis of the model, which other users should be informed about that change. In this way, all users remain up to date.
Now, as the name implies, the PDR adheres to a *distributed model of computing*. Even though various users interact through the same program (functionality as governed by the model), there is no central component, no server, no cloud and no blockchain that glues these users together.

Perspectives InPlace

This may be a unusual concept. We are so used to the client-server model that we've internalised the idea that, when interacting through computers with other people, *there is an entity in the background we are interacting with*. It's the telephone exchange model. We talk to someone, but we know there is an exchange between us. In the same vein, we know that as we use, say, Uber, we interact with a taxi driver but there is the Uber server between us. This server runs the program that connects us. It is, literally, a *connector*, like a joint in a plumbing system.
Not so with Perspectives. A Perspectives model is more like the set of rules of a board game. When we sit at the table and have the board between us, we act directly on the board and thus interact in the game context. We consider ourselves bound by the games' rules and behave accordingly. Now think of the user with his PDR as the player and the model as the game rules.

Distributed Computing

Another answer to the question "but where does the program run" is: on your and your peers devices. All these devices together are the loosely connected parts that form a machine that the program runs on. This way of computing is called peer-to-peer computing. It is safe, local and makes data centers redundant. As these centres are responsible for a large part of our collective energy consumption, peer-to-peer computing is environment-friendly to boot.
(to prevent a misunderstanding: on a low level, the exchange of information runs through large nodes in the internet. Here, the telephone exchange model is adequate. This applies, regardless whether we're talking about Uber or Perspectives. However, in the Uber case, *on top of that infrastructure* there is another centralised network. Here Perspectives is different.)