Technical Dive into AMD MLPerf Inference v6.1 Submission

AMD submitted validated MLPerf Inference v6.1 results covering Llama 2, DeepSeek R1, and GPT-OSS. The submission includes multi-node runs on 72 MI355X GPUs for the 120-billion-parameter model. These peer-reviewed benchmarks demonstrate specific hardware performance for large-scale inference workloads, providing a concrete reference point for developers evaluating AMD accelerators against other vendors.