Google DeepMindAnalysisDevelopersNovember 16, 2026

Google Cloud posts 10 MLPerf Training v6.1 results on Ironwood TPUs

Read original source →blog.google

8,192 Ironwood TPUs hit target quality in 1.70 minutes, scaling to 4,096 chips at 96% efficiency. On the first MLPerf agentic RL benchmark, 640 Ironwood TPUs reached 0.765 accuracy against a 0.69 goal in 18.02 minutes.

1 source

More stories today

Open the live feed