Search
DeepSeek Unveils New Product: The Pioneering Open-Source Model with 'Gold Medal Standard in Mathematical Olympiad'
2025-11-28 /
Read about 0 minute
Author：小编
On the evening of November 27, DeepSeek took the significant step of open-sourcing its latest model, DeepSeek-Math-V2, on the renowned Hugging Face platform. This marks a historic moment as it is the first open-source mathematical reasoning model globally to attain a performance level equivalent to a gold medalist in the International Mathematical Olympiad (IMO). Built upon the experimental architecture of DeepSeek-V3.2, this innovative model leverages a self-verification training framework. Its exceptional capabilities have been demonstrated by securing gold medals in both the IMO 2025 and CMO 2024 competitions, along with achieving an impressive score of 118 out of 120 in the Putnam 2024 test.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
