Inference & Serving · Established · Beginner

TPS

Also known as: Tokens Per Second

An industry-standard acronym standing for Tokens Per Second within inference & serving.

What TPS is

TPS (Tokens Per Second) is a fundamental terminology concept used throughout inference & serving.

How it works

Serves as shorthand in technical documentation, research papers, and engineering discussions.

Why it matters

Mastering TPS helps developers and researchers communicate effectively across the AI ecosystem.

Common uses