Chinese firm's large language model makes a splash.

Normile, Dennis · Science · 2025

editorial · Level V

Where this comes from

Abstract

DeepSeek's open-source model appears to be cheaper and faster to train and run than many others.