Inference
Running a trained model to get an answer, as opposed to training it in the first place.
Training is the one-off, enormously expensive process of building the model; inference is everything after: each question asked and each answer produced. The distinction matters for cost and for control. Training needs a data center, while inference can happen on a laptop or a phone, which is why the price of answering a query is a separate business from the price of building the model.
Where it mattered in the record
-
DeepSeek releases R1, an open-weight reasoning model
The Chinese laboratory DeepSeek released R1 on 20 January 2025, an open-weight model trained with reinforcement learning to work through problems step by step. It matched OpenAI’s o1 on several benchmarks at a small fraction of the reported training and inference cost.