Training vs. Inference
A free lesson from Swipe AI · AI Fundamentals, level 2 · updated October 2026
By the Swipe team, founded by Colin · how we make lessons
Training and inference are two separate phases, and it's easy to blur them together. Training is the (often long, expensive) process of adjusting a model's parameters using training data. Inference is what happens every single time you actually use the finished model afterward — you give it an input, and it produces an output using the parameters that training already locked in.
During training, the model's parameters are actively changing. During inference, they're fixed — using the model doesn't adjust its underlying numbers at all. Asking a trained model a question is much more like running a finished calculation than it is like teaching the model something new in that moment.
This split also explains a cost difference people often don't expect: training a large model can take a long time on powerful hardware and can be genuinely expensive, while a single inference (one response to one question) is comparatively quick and cheap — even though it's the part users actually see.
What is "inference," as distinct from training?
Every time you actually use a finished AI model to get a response, that single use is called ____.
A model's parameters keep changing every time someone uses it to answer a question.
Training Data & Parameters
10 questions · pass with 60% — just like in the app.
What is training data, in plain terms?
Modern AI models are usually trained on datasets small enough for one person to read in a single sitting.
A number inside a model that controls how strongly one piece of information influences another is called a ____.
Before training begins, what are a model's parameters typically like?
A model's parameters keep changing every time someone asks it a question.
The process of gradually adjusting a model's parameters using training data is called ____.
What is inference?
A model's entire behavior comes from the specific values of its many adjustable parameters.
A single use of an already-trained model to answer one input is called one ____.
Which best describes the cost difference between training and inference?
Free ai tools
Keep your streak going in Swipe AI
Swipe through every level of AI Fundamentals on your iPhone — with XP, stars, streaks, a personal advisor and the full track beyond these free levels.
Get the Swipe apps →