DeepSeek AI Lab Releases Updated Version of Prover AI System
Chinese AI lab DeepSeek has recently unveiled an updated version of its AI system, Prover, designed specifically for solving math-related proofs and theorems. The latest version, V2, was quietly uploaded to the AI development platform Hugging Face, showcasing significant advancements in the field of artificial intelligence.
The new Prover V2 model is built upon DeepSeek’s V3 model, boasting an impressive 671 billion parameters and adopting a sophisticated mixture-of-experts (MoE) architecture. These parameters essentially represent the model’s problem-solving capabilities, while the MoE approach breaks down tasks into smaller subtasks and delegates them to specialized “expert” components for enhanced performance.
DeepSeek’s previous update to Prover was in August, where it was hailed as a custom model for formal theorem proving and mathematical reasoning. This latest version further solidifies the lab’s commitment to innovation and excellence in the AI domain.
According to a report by South China Morning Post, DeepSeek, known for its recent release of the upgraded V3 general-purpose model, is also expected to update its R1 “reasoning” model in the near future. The lab’s continuous efforts to enhance its AI models have garnered significant attention, with reports suggesting that DeepSeek may be considering external funding for the first time.
