DeepSeek R1 — GB200 NVL72 vs H200
Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and H200 (NVIDIA Hopper) 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.
At 59 tok/s/user interactivity on DeepSeek R1, GB200 NVL72 delivers 4972 tok/s/GPU at $0.12 per million tokens; H200 delivers 422 tok/s/GPU at $0.93. GB200 NVL72 is 655% cheaper per token; GB200 NVL72 delivers 1078% more tok/s/GPU at this point.
GB200 NVL72 posts 2445 tok/s/GPU for $0.25 per million tokens at 98 tok/s/user on DeepSeek R1; H200 posts 152 tok/s/GPU for $2.66. GB200 NVL72 is 963% cheaper per token; GB200 NVL72 delivers 1507% more tok/s/GPU.
Throughput at 136 tok/s/user on DeepSeek R1: GB200 NVL72 hits 494 tok/s/GPU, H200 hits 40. Per-million costs land at $1.88 and $9.94 respectively. GB200 NVL72 is 430% cheaper per token; GB200 NVL72 delivers 1150% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | GB200 NVL72:4972.2H200:422.2 | GB200 NVL72:2444.8H200:152.1 | GB200 NVL72:494.4H200:39.6 |
| Cost ($/M tok) | GB200 NVL72:$0.123H200:$0.929 | GB200 NVL72:$0.250H200:$2.663 | GB200 NVL72:$1.876H200:$9.940 |
| tok/s/MW | GB200 NVL72:2367707H200:244043 | GB200 NVL72:1164188H200:87929 | GB200 NVL72:235423H200:22869 |
| Concurrency | GB200 NVL72:~1335H200:~133 | GB200 NVL72:~627H200:~6 | GB200 NVL72:~108H200:~14 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.