DeepSeek R1 — B200 vs H200
Head-to-head AI inference benchmark comparison of B200 (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.
Throughput at 59 tok/s/user on DeepSeek R1: B200 hits 2701 tok/s/GPU, H200 hits 422. Per-million costs land at $0.20 and $0.93 respectively. B200 is 365% cheaper per token; B200 delivers 540% more tok/s/GPU.
B200 / H200 on DeepSeek R1 at 98 tok/s/user: 521 / 152 tok/s/GPU, $1.04 / $2.66 per million tokens. B200 is 157% cheaper per token; B200 delivers 243% more tok/s/GPU.
Toward the upper edge of the 21–174 tok/s/user interactivity band, at 136 tok/s/user on DeepSeek R1: B200 runs 255 tok/s/GPU at $2.13/M tokens, H200 runs 40 at $9.94/M. B200 is 367% cheaper per token; B200 delivers 544% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:2700.6H200:422.2 | B200:521.4H200:152.1 | B200:254.9H200:39.6 |
| Cost ($/M tok) | B200:$0.200H200:$0.929 | B200:$1.037H200:$2.663 | B200:$2.127H200:$9.940 |
| tok/s/MW | B200:1244494H200:244043 | B200:240258H200:87929 | B200:117454H200:22869 |
| Concurrency | B200:~1030H200:~133 | B200:~185H200:~6 | B200:~13H200:~14 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.