Inference — Model Inference

Definition: The act of running a trained model on new input to produce output. Inference cost — not training cost — is what actually shows up on the cloud bill once a feature is live.

Example

Training was a one-time $40K — inference at scale was $40K a month until we batched and quantized.

When you'll hear it

Inference shows up most often in AI strategy reviews, model evaluation discussions, and product roadmap meetings. When someone uses it, they're usually referring to model inference — and they expect the room to already know what that means.

FAQs

What does Inference stand for?

Inference stands for Model Inference.

What does Inference mean in AI and machine-learning?

The act of running a trained model on new input to produce output. Inference cost — not training cost — is what actually shows up on the cloud bill once a feature is live.

Where will I hear Inference used at work?

Inference comes up most often in AI strategy reviews, model evaluation discussions, and product roadmap meetings. It's used as shorthand for model inference, so people assume you already know the term.