DeepSeek upgrades its math-focused AI model Prover

DeepSeek's Prover AI model has been updated with version V2 and a distilled variant, built on their advanced V3 model. This release leverages the MoE architecture to enhance problem-solving efficiency by breaking tasks into specialized subtasks. Previously updated in August as an open-source tool for formal theorem proving, DeepSeek is also exploring external funding and recently launched its general-purpose V3 model, with plans to soon update its R1 reasoning model.