China’s cheap, open AI model DeepSeek thrills scientists

A Chinese-built large language model called DeepSeek-R1 is stepping up as an affordable rival to ‘reasoning’ models such as OpenAI’s o1. R1 matched o1’s performance on certain tasks in chemistry, mathematics and coding, and has been released as ‘open-weight’, which means that researchers can study and build on the algorithm. “The openness of DeepSeek is quite remarkable,” says AI researcher Mario Krenn, since OpenAI’s models are “essentially black boxes”.