NVIDIA announced that its Vera Rubin NVL72 system debuted in the MLPerf Inference v6.1 tests, achieving up to 3.7 times higher throughput than the GB300 NVL72 in benchmarks such as DeepSeek-R1 and Qwen3-VL. Additionally, software optimizations boosted performance by up to 1.6 times compared to v6.0, with the GB300 NVL72 demonstrating 99% scaling efficiency in a 288-GPU configuration.