DeepSeek R1 — B200 vs Vera Rubin NVL72
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and Vera Rubin NVL72 (NVIDIA Vera Rubin) on DeepSeek R1. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
B200 / Vera Rubin NVL72 on DeepSeek R1 at 114 tok/s/user: 4147 / 0 tok/s/chip, $0.12 / $0.00 per million tokens. Comparable data is unavailable.
Around the middle of the 52–300 tok/s/user interactivity band, at 176 tok/s/user on DeepSeek R1: B200 runs 1912 tok/s/chip at $0.25/M tokens, Vera Rubin NVL72 runs 0 at $0.00/M. Comparable data is unavailable.
Setting 239 tok/s/user as the target on DeepSeek R1, B200 produces 1249 tok/s/chip ($0.38 per million tokens) and Vera Rubin NVL72 produces 0 ($0.00). Comparable data is unavailable. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:4146.9Vera Rubin NVL72:0.0 | B200:1912.0Vera Rubin NVL72:0.0 | B200:1249.5Vera Rubin NVL72:0.0 |
| Cost ($/M tok) | B200:$0.116Vera Rubin NVL72:$0.000 | B200:$0.251Vera Rubin NVL72:$0.000 | B200:$0.385Vera Rubin NVL72:$0.000 |
| tok/s/MW | B200:2425080Vera Rubin NVL72:0 | B200:1118156Vera Rubin NVL72:0 | B200:730700Vera Rubin NVL72:0 |
| Concurrency | B200:~52Vera Rubin NVL72:~400 | B200:~28Vera Rubin NVL72:~400 | B200:~19Vera Rubin NVL72:~400 |