newsnewsJul 29, 2026DeepSeek-R1 Benchmarks: Open Reasoning Shocks IndustryDeepSeek releases R1 open-weights model matching proprietary reasoning performance on math and coding.