MLCommons releases Training v6.1 with first Rubin results

MLCommons released MLPerf Training v6.1 on 16 Nov 2026, with 153 systems and 250 results, and said the round included Nvidia Rubin-generation GPUs and Google's TPUv7, called Ironwood, for the first time. The results explorer lists Nvidia's Hecate Rubin system at 3.947 minutes on Llama 3.1 8B with 512 GPUs and a CoreWeave Rubin system at 7.162 minutes with 72 GPUs. Nvidia says Vera Rubin NVL72 delivered up to twice the performance per GPU of its GB300 NVL72.

/ Sources

/ About this story

Compiled by Venture Atlas from the sources above, using automated AI-assisted research. This is a summary of reporting published elsewhere, not original reporting - follow the source links for the full account. See our editorial standards.

Something wrong here? Email apostructura@gmail.com and we will fix it.

/ Related