NVIDIA Vera Rubin NVL72 Delivers Leading Performance in MLPerf Inference v6.1 Debut
NVIDIA's Vera Rubin NVL72 tops MLPerf Inference v6.1 — the first public benchmark showing where inference performance stands as token economics reshape AI unit economics.
Why it matters
Inference performance benchmarks are the practitioner's lens into real-world token-serving costs. MLPerf v6.1 is the standard measure; NVIDIA leading it signals the inference hardware roadmap and validates scaling assumptions that drive AI capex and margin models.
The key facts
9 to knowNVIDIA Vera Rubin NVL72 leads MLPerf Inference v6.1
System performance directly tied to tokens-per-second and revenue per deployment
Efficient scaling cited as key lever for proportional throughput gains
Continuous software optimization emphasized as value multiplier on existing hardware
Published September 16, 2026 on official NVIDIA blog
NVIDIA Vera Rubin NVL72 system debuts in MLPerf Inference v6.1
Benchmark: system performance, scaling efficiency, and software optimization as inference economics drivers
Focus: tokens/cost, throughput scaling linearity, infrastructure ROI
Published Sep 16 2026 on NVIDIA blog (vendor-sourced)
Go to the source
NVIDIA Blogblogs.nvidia.com
Publisher excerpt: System performance, efficient infrastructure scaling and continuous software optimization are key levers that determine AI inference economics. Higher system performance means more tokens generated, resulting in higher revenue. Efficient scaling means throughput grows proportionally as hardware…