The internal numerical parameters a neural network learns during training, which determine how it processes input to produce output.
A model's weights encode everything it 'knows.' Fine-tuning adjusts these weights; prompting does not — it works entirely within the model's existing, fixed weights.
