AMD released MLPerf Inference v6.1 results on September 16, 2026, saying it achieved leadership scores across multiple benchmarks. It is the company's fifth consecutive MLPerf Inference update since its first appearance in the benchmarking series.
AMD reported leadership performance with AMD Instinct™ MI355X and MI350P GPUs delivering top results across multiple benchmarks, measured on the dlrm-v3 and llama2-70b models. That compares with prior rounds where AMD expanded its footprint over prior rounds, spanning recommendation, multimodal, reasoning, and large-language-model inference.
The MLPerf Inference v6.1 results are built on the CDNA 4 architecture and target enterprise generative and agentic AI workloads. Availability begins with validated results across five models, initially for customers using AMD Instinct GPUs.
"This round introduces the AMD Instinct™ MI350P, a PCIe-based GPU for enterprise inference that posted the top result among all PCIe-based submissions," said Meena Arunachalam, Miro Hodak. The results are fully reproducible and readers can do their own measurements by following step-by-step instructions in our reproduction blog.
The announcement follows AMD's fifth consecutive MLPerf Inference round, reflecting its sustained investment in transparent, standards-based benchmarking. [Company] did not say how the results will be applied beyond the current benchmarks, and [the open question or limitation the source raises] is whether the performance gains can be replicated in real-world deployments. [One sentence on what the source says happens next].
Source: amd
