Definition
Running a trained model on input to produce an output.
Distinguish it from nearby terms
Inference is model execution, not the training process and not necessarily logical proof.
Check your understanding
Agent loops may perform many inference calls plus tool actions before producing one outcome.