DeepSeek R1 — H100 vs H200
Head-to-head AI inference benchmark comparison of H100 (NVIDIA Hopper) 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.
H100 / H200 on DeepSeek R1 at 45 tok/s/user: 739 / 1779 tok/s/GPU, $0.56 / $0.22 per million tokens. H200 is 154% cheaper per token; H200 delivers 141% more tok/s/GPU.
Around the middle of the 20–122 tok/s/user interactivity band, at 71 tok/s/user on DeepSeek R1: H100 runs 281 tok/s/GPU at $1.37/M tokens, H200 runs 904 at $0.43/M. H200 is 221% cheaper per token; H200 delivers 222% more tok/s/GPU.
Setting 97 tok/s/user as the target on DeepSeek R1, H100 produces 154 tok/s/GPU ($2.45 per million tokens) and H200 produces 500 ($0.78). H200 is 213% cheaper per token; H200 delivers 224% more tok/s/GPU. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H100:739.5H200:1778.8 | H100:280.6H200:903.7 | H100:154.3H200:500.1 |
| Cost ($/M tok) | H100:$0.557H200:$0.220 | H100:$1.366H200:$0.426 | H100:$2.453H200:$0.784 |
| tok/s/MW | H100:427454H200:1028222 | H100:162224H200:522364 | H100:89204H200:289088 |
| Concurrency | H100:~132H200:~115 | H100:~41H200:~122 | H100:~17H200:~4 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.