DeepSeek

DeepSeek is a Chinese company that develops large language models. In late January 2025 it astounded the world by claiming its R1 model was trained with vastly fewer resources than rivals’ products. That claim saw share prices slump at companies such as Nvidia and Broadcom that produce hardware for AI. Analysis of the methods DeepSeek had used to train its models suggest it did not produce them from scratch and instead leant on work by OpenAI.