AI glossary
Open-weights model
A model whose trained weights are published, so anyone can download it and run or adapt it on their own hardware.
With an open-weights model you are not limited to the provider’s API: you can run it locally or on your own servers, which helps with privacy, cost control and customisation through fine-tuning. Licences vary and some restrict commercial use, so read them.
“Open weights” is not always the same as “open source”: the training data and code are often not released.
Example: Mistral releases Mistral Small 4 and Mistral Large 3 under the Apache 2.0 licence, which allows commercial use, while Mistral Medium 3.5 uses a modified MIT licence. Two “open” models from the same company can come with different terms.
In practice
- Read the licence before using it in a product.
- Check what hardware you need; quantization helps it fit on modest machines.
- Running it locally avoids sending data to third parties, but security and updates become your job.
Who releases what, in our AI model guide.
