Back to sectionEvent · 2025Open modelsDeepSeek-V3 and R1Strong reasoning at a radically lower training cost.SourcesDeepSeek-V3 · technical reportOpen primary source