MLCommons Releases MLPerf Training v6.0 Results

MLCommons released MLPerf Training v6.0 benchmark results featuring two new benchmarks (DeepSeek V3 and GPT-OSS 20B) emphasizing sparse computation and Mixture-of-Experts architecture. The round achieved record participation with 95 unique systems from 24 organizations, demonstrating growing diversity in AI training hardware, accelerators, and cloud-based solutions.