Parameter
One adjustable number inside a model, learned in training. Model size is usually given as a parameter count.
Also: parameters, parameter count
Parameters are the weights and biases of a neural network, the numbers training adjusts. Their count is the usual measure of a model's size: BERT-large has 340 million, GPT-3 has 175 billion, Llama 3.1's largest model has 405 billion, and Kimi K2 about one trillion. More parameters can store more knowledge and skill, given enough data and compute, but also cost more memory and power to run. Many frontier labs no longer publish parameter counts for their closed models.