Menu

Artificial Intelligence

Inference

Plain English

When a produces an answer.

Definition

is the of using a trained to generate outputs or make predictions based on new input data.

In practice

Occurs when users interact with AI , such as generating text or predictions.

In context

It is the cost that recurs. Training is a large one-off; is charged on every , which is what makes usage a commercial decision as much as a technical one.

The reality

can be resource-intensive and may vary in speed and cost depending on the .

Compared with

Inference vs Training

Training the from and happens rarely. Inference uses it to produce an output and happens constantly. Training dominates the initial cost; inference dominates the running one.

FAQ

Common questions

A few practical answers to the questions that usually come up around this term.

What is inference in AI?

It is the of generating outputs using a trained .

When does inference happen?

When a is used to respond to new input.

Why is inference important?

It is how deliver value in real-world use.

What affects inference performance?

size, , and input complexity.

Related Services

Related Guides

Related Terms

LET'S WORK TOGETHER

Ready to improve your product?

UX, research and product leadership for teams tackling complex digital services.

Previous feedback

I had a fantastic experience working with Andy. One of his most impressive achievements during our time at NHS HEE was masterminding a deeply complex information architecture for a new platform that brought together a large number of legacy websites.

Will Parkhouse

Senior Content Designer