TMTPOST -- Chinese start-up DeepSeek quietly open-sourced a new specialist artificial intelligence (AI) model on Wednesday, just one day after Alibaba unveiled the third generation of its Qwen model family, underscoring the intensifying competition in generative AI development. The Hangzhou-based company uploaded its latest Prover-V2 model to Hugging Face, the world’s largest open-source AI community platform, without issuing any announcements on its official social media channels. The release comes amid mounting anticipation for DeepSeek’s upcoming R2 reasoning model, which is expected to launch in the near future. The Prover series comprises domain-specific models tailored for solving mathematics-related problems. While DeepSeek has yet to publish formal documentation for the new model on Hugging Face, files reviewed by the Post indicate that Prover-V2 is built on DeepSeek’s V3 base model, which features 671 billion parameters and leverages a mixture-of-experts architecture aimed at optimizing training efficiency and operational costs.