In MLPerf Inference v6.1 results, Crusoe reported 5.75 million tokens per second for gpt-oss-120b using 512 AMD Instinct MI355X GPUs.
Crusoe reports MLPerf Inference v6.1 results from a cluster with 512 AMD Instinct MI355X GPUs: 5.75 million tokens per second in the gpt-oss-120b test. The company also reported linear scalability without InfiniBand.
These figures offer a reference for comparing throughput and scalability across inference clusters, but do not guarantee equivalent performance with other workloads or configurations. When using AI to study or apply these results, avoid entering personal or internal data unless necessary, and check the data-handling policies of the tool you use.