Model Weights
Also called Weights
The numbers inside a trained AI model that store what it has learned. Sharing a model's weights means sharing the model itself.
What are model weights?
Model weights are the numeric values in a neural network that determine how it turns an input into an output. Training adjusts them until the model gives good answers. After training, the weights are the learned knowledge of the model, stored as a large file.
How do they work?
Each weight scales how strongly one part of the network influences another. A model with billions of parameters has billions of these values. During training they are tuned automatically; during use they stay fixed. Because they are plain numbers, they can be copied, downloaded and run on suitable hardware.
What is a simple example?
When a lab releases an open-weight model, it publishes this file so anyone can run the model on their own machine or fine-tune it. A closed model keeps its weights private and can only be used through the provider's service.
Why does it matter?
Whoever holds the weights controls the model: who can run it, change it or inspect it. That is why weights are central to debates about open and closed AI.