Learn AI / Compare /Training vs inference

Training vs inference

Is the model changing, or just being used?

The short answer

Training is the expensive, occasional process that sets the model's internal numbers. Inference is what happens every time you use it afterwards. Asking a chat assistant a question does not train it, and your question does not become part of the model.

Training

How Machines Learn · Starter

In one sentence: Training is the process of adjusting a model's parameters using data or feedback to improve performance on a learning objective.
Why it matters: If you hear that a system was trained, you should ask what data shaped it, what feedback it used, and whether it was tested on new cases. Those details affect accuracy, fairness, safety and whether the system fits the task at all.
The everyday version: Training means showing a model many examples and adjusting it so its future outputs are more useful.
The one thing to remember: Training adjusts a model’s parameters using examples and a learning objective. Good results on training data alone do not show that the model will generalise to new inputs.
The commonest misconception: Training means typing instructions into a system. A prompt gives directions at use time, but training changes the model itself using many examples or feedback signals over repeated updates.

Read the full page

Inference

How Machines Learn · Starter

In one sentence: Using a trained model on new input to produce an output such as a score, category, translation or generated text.
Why it matters: If you use AI tools, inference is the moment where a result is actually produced. Knowing this helps you ask practical questions: How fast is it, what input does it need, what kind of output will it give, and what human checking is still necessary before action is taken?
The everyday version: Inference is using an already trained AI system on new data to get a result.
The one thing to remember: Training establishes the model's learned parameters. Inference applies them to new input to produce an output.
The commonest misconception: Inference means the model is still learning from each new user input. Usually it is not learning during this step. In standard use, the model's settings were set earlier in training and are simply applied to new input.

Read the full page