Models & architecture
Model parameters
The learned numerical values that determine a model's behaviour.
Also known as: weights
Definition
Parameters are weights adjusted during training that encode the statistical patterns a neural network uses during inference.
Why it matters
Parameter count is an imperfect proxy for capability and says nothing by itself about data quality, architecture or operational efficiency.
Related concepts
- Neural network
A layered mathematical model that learns relationships between inputs and outputs.
- Open-weight model
A model whose trained parameters are available to download or host.
- Model training
The process of adjusting model parameters using examples and an objective.