Glossary Entry

Inference

The stage where a trained model is used to make predictions or generate outputs on new inputs.

Deployment Models

Seed source: Google ML Glossary

Inference is the part of the lifecycle that most users actually experience. Training builds the model, but inference is where the model classifies, ranks, generates, or recommends something in response to live data.

Latency, cost, and reliability matter heavily here, because even a strong model can feel poor if the inference path is slow or brittle.