DeepSeek Releases New Mathematical Model for AI: Prover-V2

Chinese startup DeepSeek has silently released its latest model, Prover-V2, on Hugging Face. The move follows a trend of other companies releasing their own advancements in AI, including Alibaba’s Qwen3, which claims to outperform DeepSeek’s R1 and OpenAI’s o1 reasoning models. DeepSeek is known for its expertise in mathematical problem solving through its Prover series models and has not yet detailed the specifics of Prover-V2 on its website. However, analysis suggests it builds on the V3 model with an impressive 671 billion parameters.