Google Cloud posts MLPerf Training v6.1 benchmark results
Google
· November 16, 2026
· ✓ verified
Google Cloud has published 10 MLPerf Training v6.1 benchmark results across its seventh-generation Ironwood TPUs and NVIDIA Grace Blackwell GPUs.
- The results include frontier pre-training and the first MLPerf benchmark for agentic reinforcement learning (RL), with Ironwood TPUs reaching target quality in 1.70 minutes at 8,192 chips and 96% efficiency at 4,096 chips.
- On A4X Max Instances, 256 NVIDIA GB300 GPUs reached target quality in 31.96 minutes; Google Cloud also reported five A4X and A4X Max submissions on the agentic RL benchmark, and linked the results to its AI Hypercomputer architecture.