Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

MLPerf Training v6.1 spotlights Vera Rubin and Ironwood

AI summary

MLPerf Training v6.1 introduced NVIDIA’s Vera Rubin architecture and the first agentic reinforcement-learning benchmark, with Google Cloud, NVIDIA and other partners submitting results. NVIDIA said Vera Rubin NVL72 achieved the fastest training time on six of seven workloads and up to twice the performance of an optimized GB300 NVL72 baseline, while a 10,240-GPU GB300 cluster posted the fastest absolute DeepSeek-V3 671B result. Google Cloud reported that 8,192 Ironwood TPUs completed DeepSeek-V3 pretraining in about 102 seconds and that 640 Ironwood chips completed the Qwen3.5-397B agentic RL test in 18.02 minutes with 76.5% accuracy; these performance comparisons remain vendor-reported.

In-depth insight

The benchmark round provides a new multi-vendor view of large-scale AI training and agentic reinforcement-learning performance, although the headline comparisons have not been independently verified.

As AI agents tackle increasingly complex workflows, reducing time-to-train is critical for rapidly iterating on both models and the agents they power. MLPerf Training v6.1 marks the debut of the NVIDIA Vera Rubin architecture, and these early results demonstrate that its CUDA software stack is already highly mature. NVIDIA swept all seven benchmarks, with the Vera Rubin NVL72 platform delivering the fastest time to train on six of the seven, achieving up to 2x more performance than a highly optimized NVIDIA GB300 NVL72 baseline.

For the seventh benchmark, DeepSeek-V3 671B, NVIDIA captured the absolute fastest time of the round by demonstrating massive scale on a 10,240-GPU GB300 NVL72 cluster. This dual-platform approach deliberately showcases the strengths of both systems. While the GB300 results highlight the massive, reliable cluster scale NVIDIA is deploying today, Vera Rubin demonstrated its own exceptional scaling architecture by maintaining above 97% overall efficiency when scaling from 256 to 1,024 GPUs. Together, NVIDIA platforms hold the fastest time to train on every benchmark in v6.1.

NVIDIA platforms were also the only ones to submit results across all seven benchmarks, with Vera Rubin successfully completing the entire suite in its inaugural run. For developers, this breadth matters as much as any single top-line number. It proves that the performance gains inherited on day one span the entire modeling pipeline, comprehensively supporting everything from large-scale pretraining to the reinforcement learning post-training required for modern AI agents.

#NVIDIA#Google Cloud#MLCommons#Vera Rubin NVL72#GB300 NVL72#Ironwood TPU#Grace Blackwell GPU#DeepSeek-V3 671B#Qwen3.5-397B

Original · 10 min readRead the original
Group chat

No comments yet. Start the conversation.