Inference
Running a trained model to get a prediction — the 'using it' phase, as opposed to training. Every time you send a prompt and get an answer back, that's one inference.
Think of It Like This
Training is studying for the exam; inference is sitting it.
During inference the weights are frozen — the model just runs the input forward and produces an output. For language models that means generating one token at a time, feeding each new token back in to predict the next. Inference is where a deployed model's ongoing cost and latency live, so a lot of engineering goes into making it fast and cheap: batching requests, caching, and quantizing the weights.