Open or closed model: what do you actually get?
“Open” can mean open code, downloadable weights, known training data or free use — and these are not the same thing.
What is it?
An “open” model can mean very different things: open source code, freely available weights, known training data or simply free use. A closed model is available only through the provider’s service. For any model it is worth checking what exactly is open and under what licence.
An everyday example
A company can install an open-weight model on its own server and use it without sending data to an outside provider, but the licence may still restrict commercial use.
Why does it matter?
Openness affects privacy, costs, dependence on a provider and whether the model can be checked independently.
A common misconception
That “open source”, “open weights” and “free” mean the same thing.
In more detail
What are open weights?
The provider lets you download the trained numerical parameters. That does not mean it also publishes the training data or the training code.
What is a closed model?
A model you mainly access through the provider or its service, without access to all of its internal components.
What is the advantage of running it locally?
More control over the environment and sometimes over privacy. But you still have to protect the device, the data and the permissions.
What should I check before choosing?
The licence, usage restrictions, where the data is processed, quality in your language, costs, energy use and security.
Remember: “Open source”, “open weights” and “free” are not synonyms.
Reliable sources
Last reviewed 1 November 2026. An explainer, not professional, legal or medical advice. For rules and figures that change, check the official version.