Models & Architecture
Parameters
The adjustable numbers inside a model that get tuned during training and store what the model has learned.
Parameters are the internal values, mostly the connection weights, that a model adjusts as it trains. A models size is usually quoted as its parameter count, from millions to hundreds of billions, because more parameters generally mean more capacity to capture patterns. Parameter count is a rough proxy for capability, but it is not the whole story, since data quality and training method matter just as much.
Related terms
Weights
The individual strengths of the connections in a neural network, which are what actually change as a model learns.
Neural Network
A model loosely inspired by the brain, built from layers of simple units that each pass signals to the next.
Large Language Model
A model trained on massive amounts of text that predicts and generates language, powering tools like ChatGPT and Claude.
Training Data
The examples a model learns from, whose size, quality, and bias largely determine how the model turns out.