DeepSeek Says Upgraded Model Reasons Better, Hallucinates Less

DeepSeek’s update was announced on Wednesday.

Photographer: Andrey Rudakov/Bloomberg

The Chinese startup DeepSeek said Thursday that its upgraded artificial-intelligence model can perform mathematics, programming, and general logic better than the previous version, while hallucinating less.

The upgrade to its R1 model — which stunned the AI world in January by rivaling the systems of much-larger US developers despite being built at what the Chinese startup said was a fraction of the cost — features a greater depth of reasoning, DeepSeek said in a post on the AI model platform Hugging Face.