What are open AI models?
Open models are AI models whose weights have been published, so that others can download them and run them in their own environment.
A closed model can only be used through the supplier's service. An open model can run on your own servers, even without internet.
Open means different things. Often only the model's weights are free to use, while the training data and the code that produced it are not. Licences differ too, and some restrict commercial use.
Open models have improved a great deal in recent years. For many tasks in an organisation they are more than enough, especially when they read the right documents.
How it works
The weights are the numbers the model learned during training. Once they are published, anyone with enough hardware can download them and run the model. They can also be fine-tuned or combined with your own documents, within the terms of the licence.
Why it matters
Open models make it possible to use AI without data leaving the organisation and without being tied to a single provider's service. The model can be swapped when a better one comes along, and cost depends on your own operation.
Open models can also be examined freely. Researchers and users can test them independently, and known weaknesses are often described publicly within a short time.
What to keep in mind
An open model comes as it is. There is rarely anyone responsible for how it behaves. Test it on your own tasks, in the languages you work in, and keep track of new versions.
Licences can also change between versions. Read the terms for the version you actually use.
What it means for you
Read the licence before you choose an open model, and check that it works well in Swedish.
Common questions
Are open models the same as open source?
Not always. Often only the weights are published, while training data and code are not. The licence decides what you may do with the model.
Are open models safe to use?
Safety depends mainly on where and how the model runs. An open model in your own environment sends no data anywhere. Review the licence and where the model comes from.
Why do organisations choose open models?
They can run in the organisation's own environment, even fully offline. That gives control over data and less dependence on a single supplier.
Related terms
Read more
All terms Updated