Inference
Using a fully trained model to compute an output for a new input.
Every time you ask a chat assistant a question, inference takes place: the model computes an answer from your input. The model does not learn anything new in the process – that only happens during training.
Inference requires computing power, usually on GPUs. Cost and response time depend on model size, the length of input and output, and the hardware.
General information, not legal or professional advice.