Understand this story
Google Cloud announces 10 MLPerf Training v6.1 results, covering the seventh-generation Ironwood TPU and NVIDIA Grace Blackwell GPUs. Among them, 8,192 Ironwood TPUs reached the DeepSeek-V3 benchmark target quality in 1.70 minutes, with 96% efficiency when scaled to 4,096 units; 640 Ironwood TPUs achieved a target accuracy of 0.765 (target 0.69) in 18.02 minutes on the first agentic reinforcement learning benchmark. On A4X Max instances, 256 NVIDIA GB300 GPUs reached target quality in 31.96 minutes.
AI-generated from public reports
Latest development · Report
Google Cloud announces MLPerf Training v6.1 results: Ironwood TPU and…Developments
1 confirmed developments
- Source:Google Blog
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 16
Google BlogFirst-party
Google Cloud announces MLPerf Training v6.1 results: Ironwood TPU and…
Google Cloud announced 10 results in MLPerf Training v6.1, covering the seventh-generation Ironwood TPU and NVIDIA Grace Blackwell GPUs. 8,192 Ironwood TPUs reached the target quality of the DeepSeek-V3 benchmark in 1.70 minutes, with 96% efficiency when scaled to 4,096 chips; 640 Ironwood TPUs reached a target accuracy of 0.765 (target 0.69) in 18.02 minutes on the first agentic reinforcement learning benchmark. On A4X Max instances, 256 NVIDIA GB300 GPUs reached the target quality in 31.96 minutes.