MLPerf Inference v6.1: 5.7x per-accelerator gains, 512-GPU run, Vera Rubin results

MLCommons published MLPerf Inference v6.1: 30 participating organizations and 486 results. End-to-End RAG and Edge Agentic Inference tests were added, and results from NVIDIA Vera Rubin NVL72, AMD Instinct MI350P, Intel Arc Pro B70 and Ryzen AI Max+ 395 passed review for the first time.

• Best DeepSeek-R1 result per accelerator grew 5.7x year over year
• VLM test improved 2.99x in half a year since v6.0
• Crusoe and AMD showed the largest system in MLPerf history — 512 GPU MI355X
• NVIDIA claims up to 2.5x token gain for Vera Rubin NVL72 vs GB300