In plain English
Parameters, also called weights, are the internal numerical values a model adjusts during training to capture patterns in data. A large model can have billions of them, and together they encode everything the model has learned. After training they are fixed, and the parameter count is a rough indicator of a model's capacity.
Why it matters
Parameter counts are widely quoted as a measure of how powerful a model is, so it helps to know they indicate scale rather than guaranteed quality.
A worked example
When people say a model 'has 70 billion parameters', they mean 70 billion tunable values were adjusted during training to encode its abilities.
Common confusion
More parameters do not simply mean a better model. A smaller, well-trained model often beats a larger, poorly trained one on a given task.