AMD · News · 20260916N

AI Inference Product Performance

ADVANCED MICRO DEVICES INC · 2026-09-16 · Importance 23 · Surprise 24

AMD’s MLPerf Inference 6.1 submission covered its Instinct MI355X and MI350P GPUs. The company reported record aggregate token throughput with partners and performance gains from software optimization on unchanged hardware. The results demonstrate scalability and production-oriented inference capabilities across multiple deployment requirements.

Key facts

• AMD delivered its broadest MLPerf Inference 6.1 submission. source
• AMD's MLPerf Inference 6.1 submission demonstrates advancements in AI inference performance and scalability across its Instinct MI355X and MI350P GPUs. source
• AMD achieved record aggregate token throughput with its partners in its MLPerf Inference 6.1 submission. source
• AMD showed substantial performance improvements from the same hardware through continuous software optimization in its MLPerf Inference 6.1 submission. source