Parameters
The numbers inside a model that training adjusts; their count, in billions, is the usual measure of a model's size.
Also: weights, model size
A model's parameters, or weights, are the values that determine its output. GPT-3, in 2020, had 175 billion; leading open models of 2025 range from a few billion to over a trillion. Each parameter takes 2 bytes at 16-bit precision, so parameter count sets how much GPU memory a model needs: a 405-billion-parameter model needs about 810 GB just for its weights at that precision, more than one eight-GPU H100 server holds. In mixture-of-experts models, only some parameters are used for each token.